LLM benchmark results, 2026-08-13

Metric Key

Quick reference for the columns in every table below.

Metric	Range	Direction	What it means
F1 (accuracy)	0 to 1	higher is better	Combined score of precision and recall against the human-verified ground-truth ad spans. F1 = 0 means the model found nothing right; F1 = 1 means it found every ad with the correct boundaries. Uses IoU >= 0.5 (predicted span must overlap truth span by at least half) to count a match.
Cost / episode	USD	lower is better	Average dollars per episode at the current pricing snapshot. Recomputed from token counts so all rows compare at the same prices regardless of when the call ran.
F1 / $	ratio	higher is better	F1 divided by cost-per-episode. Cheap accurate models score highest. Free-tier models (when the roster has any) are rank-listed separately because the ratio is undefined.
p50 / p95 latency	seconds	lower is better, with caveats	Median (p50) and tail (p95) wall-clock response time. Note: for models routed through OpenRouter (everything except claude-*), this includes OpenRouter’s queueing and upstream-provider latency, not just the model itself. Treat as a load/availability indicator, not a model-quality signal.
JSON compliance	0 to 1	higher is better	Fraction of responses that parsed as a clean JSON array matching the requested schema. 1.0 = always clean; lower = used object wrappers ({ads: [...]}), markdown fences, extra fields like sponsor, or required regex fallback to extract.
No-ad episode	PASS / FAIL	PASS desired	Negative-control test on the 2 episode(s) verified to contain no ads (ep-ai-cloud-essentials-e8dc897fbd6b, ep-oxide-and-friends-ce789ff5b62e). PASS = zero predictions across all 16 of their windows. FAIL = the model false-positived on a non-ad segment, with the FP count shown.
F1 stdev	0 to 1	lower means more consistent	Standard deviation of F1 across the 12 ad-bearing episodes. High stdev = inconsistent across content types.
Moderation blocked	0 to 100%	0 desired	Share of attempted calls the provider refused on content grounds. Refused windows never reach scoring, so any non-zero value means that model’s F1 was computed on a subset of the corpus and is not comparable to an unblocked row.
JSON mode	native / prompt-inject / mixed	–	How the model received its JSON-output instruction. native = provider accepted response_format=json_object for at least 95% of calls; prompt-inject = provider rejected it and the runner fell back to instructing JSON in the prompt for at least 95% of calls; mixed = neither path crossed the threshold (sample mostly comes from intermittent provider rejections). Reads from json_format_used in calls.jsonl. Useful when picking a model whose provider may not support native JSON mode – a strong JSON compliance score from a prompt-inject model carries different weight than the same score from a native model.

Glossary

• IoU (intersection over union): how much two time ranges overlap, expressed as (overlap) / (union). 0 means no overlap, 1 means identical ranges. We use IoU >= 0.5 as the threshold for a predicted ad to count as matching a truth ad.
• Trial: each (model, episode) pair runs 5 trials at temperature 0.0 to surface non-determinism. F1 numbers in tables are averaged across trials.
• Window: each episode is split into ~85-second sliding windows; the model judges each window independently. Per-window predictions are stitched together for episode-level scoring.
• Schema violations: number of times the response had at least one missing-required-field, wrong-type, or extra-key issue. Doesn’t tank F1, but signals brittleness.
• Extraction method: the route the parser took to recover the ad list. json_array_direct is the cleanest; method names with regex_* mean the JSON itself was malformed and we fell back to text matching.

TL;DR

Best Accuracy (F0.5 @ IoU >= 0.5)

Models ranked by F0.5 (precision weighted 2x recall) against human-verified ground truth. MinusPod cuts the segments it flags, so cutting real content (a false positive) is worse than leaving an ad in (a false negative), and F0.5 penalizes it more. A model shares the tier above it unless it scores consistently lower across the same episodes (paired one-sided t-test, 95%); models that trade wins episode to episode share a tier, so order within a tier is not meaningful on this 12-episode corpus. Flags caveat a model without changing its rank.

Tier	Model	F0.5	95% CI	Precision	Recall	F1	Cost / episode	p50 latency	JSON compliance	Flags
A	claude-haiku-4-5-20251001	0.819	+/-0.115	0.807	0.899	0.842	$1.0820	24.2s	1.00
A	claude-sonnet-4-6	0.791	+/-0.115	0.795	0.846	0.799	$3.2386	4.1s	1.00
A	google/gemini-3.6-flash	0.788	+/-0.122	0.788	0.840	0.796	$2.8863	4.5s	1.00
A	x-ai/grok-4.5	0.781	+/-0.121	0.777	0.853	0.797	$3.3554	10.9s	1.00
A	openai/gpt-5.5	0.780	+/-0.124	0.783	0.819	0.784	$7.6842	5.4s	0.87	(!) brittle JSON (!) fails no-ad control
A	google/gemini-3.5-flash	0.774	+/-0.124	0.771	0.838	0.787	$3.4202	5.6s	1.00
A	google/gemini-3.5-flash-lite	0.774	+/-0.110	0.786	0.813	0.776	$0.3643	0.7s	1.00
A	openai/gpt-5.6-terra	0.767	+/-0.126	0.772	0.808	0.768	$1.1951	2.1s	0.86	(!) brittle JSON (!) fails no-ad control
A	x-ai/grok-4.3	0.767	+/-0.122	0.755	0.847	0.789	$1.6048	3.7s	1.00
A	qwen/qwen3.5-plus-02-15	0.756	+/-0.115	0.745	0.844	0.779	$1.0746	29.2s	1.00
A	claude-fable-5	0.752	+/-0.112	0.740	0.878	0.783	$10.7552	6.6s	1.00
A	mistralai/mistral-medium-3-5	0.746	+/-0.104	0.724	0.886	0.787	$1.8393	1.1s	1.00
A	qwen/qwen3.6-flash	0.738	+/-0.088	0.737	0.788	0.747	$0.5435	7.6s	0.96
A	openai/gpt-5.6-luna	0.736	+/-0.119	0.730	0.824	0.754	$0.1402	3.5s	0.83	(!) brittle JSON
A	google/gemini-3.1-pro-preview	0.732	+/-0.119	0.726	0.826	0.752	$4.9160	8.3s	1.00
A	claude-opus-4-8	0.728	+/-0.101	0.715	0.857	0.760	$5.3661	7.8s	1.00
A	claude-opus-5	0.725	+/-0.143	0.718	0.845	0.748	$5.3761	3.7s	1.00
B	deepseek/deepseek-v4-flash	0.723	+/-0.133	0.746	0.717	0.708	$0.2046	6.3s	0.82	(!) brittle JSON
B	qwen/qwen3.7-max	0.718	+/-0.121	0.706	0.826	0.746	$2.8109	23.4s	0.99
B	openai/gpt-5.6-sol	0.716	+/-0.135	0.709	0.816	0.738	$6.2293	3.7s	0.84	(!) brittle JSON (!) fails no-ad control
B	claude-sonnet-5	0.708	+/-0.128	0.698	0.812	0.732	$2.1452	8.5s	1.00
B	deepseek/deepseek-v4-pro	0.708	+/-0.133	0.739	0.665	0.680	$0.6692	16.2s	0.87	(!) brittle JSON (!) fails no-ad control
B	claude-opus-4-7	0.704	+/-0.129	0.693	0.816	0.731	$5.3293	4.6s	1.00
B	qwen/qwen3.7-flash	0.701	+/-0.093	0.702	0.740	0.706	$0.0725	10.1s	0.96
B	moonshotai/kimi-k3	0.693	+/-0.130	0.694	0.756	0.703	$5.9242	13.8s	0.81	(!) brittle JSON (!) fails no-ad control
B	deepseek/deepseek-v3.2	0.687	+/-0.118	0.792	0.552	0.599	$0.3015	1.7s	1.00
B	google/gemini-3.1-flash-lite	0.686	+/-0.108	0.653	0.918	0.747	$0.2988	0.8s	0.94	(!) fails no-ad control
B	meta/muse-spark-1.1	0.671	+/-0.128	0.765	0.530	0.595	$2.2536	4.3s	0.93
B	google/gemini-2.5-pro	0.669	+/-0.132	0.656	0.794	0.700	$4.1601	14.2s	0.96	(!) fails no-ad control
B	mistralai/mistral-medium-3.1	0.669	+/-0.136	0.655	0.790	0.699	$0.4847	0.7s	1.00
B	minimax/minimax-m3	0.665	+/-0.083	0.720	0.612	0.626	$0.3575	1.7s	0.88	(!) brittle JSON (!) fails no-ad control
B	google/gemma-4-31b-it	0.662	+/-0.132	0.665	0.701	0.667	$0.1181	2.4s	0.87	(!) brittle JSON (!) fails no-ad control
C	deepseek/deepseek-v4-flash-0731	0.656	+/-0.148	0.720	0.557	0.601	$0.1482	13.2s	0.73	(!) brittle JSON
C	openai/gpt-oss-120b	0.655	+/-0.146	0.636	0.807	0.692	$0.0634	6.5s	0.88	(!) brittle JSON (!) fails no-ad control
C	google/gemini-2.5-flash	0.641	+/-0.156	0.612	0.859	0.694	$0.3796	0.8s	1.00
C	qwen/qwen3.6-plus	0.622	+/-0.166	0.645	0.619	0.607	$1.1534	36.9s	0.90	(!) brittle JSON
C	qwen/qwen3.5-27b	0.607	+/-0.096	0.675	0.563	0.567	$0.8941	37.6s	0.70	(!) brittle JSON
C	nvidia/nemotron-3-super-120b-a12b	0.597	+/-0.209	0.624	0.577	0.577	$0.2194	22.3s	0.79	(!) brittle JSON (!) fails no-ad control
C	openai/gpt-5.4	0.592	+/-0.116	0.574	0.776	0.634	$2.7907	1.4s	0.80	(!) brittle JSON (!) fails no-ad control
C	z-ai/glm-5.2	0.568	+/-0.121	0.555	0.769	0.606	$0.7376	3.5s	0.73	(!) brittle JSON (!) fails no-ad control
C	google/gemini-2.5-flash-lite	0.566	+/-0.153	0.540	0.854	0.623	$0.1132	0.8s	0.97	(!) fails no-ad control
C	openai/o3	0.563	+/-0.173	0.710	0.371	0.457	$3.3100	6.7s	0.93
C	deepseek/deepseek-r1	0.552	+/-0.151	0.546	0.674	0.574	$1.4743	35.7s	0.84	(!) brittle JSON (!) fails no-ad control
C	tencent/hy3	0.548	+/-0.221	0.615	0.441	0.489	$0.3876	22.3s	0.62	(!) brittle JSON
D	google/gemma-4-26b-a4b-it	0.545	+/-0.169	0.532	0.651	0.572	$0.0825	1.4s	0.84	(!) brittle JSON (!) fails no-ad control
D	openai/gpt-5.4-mini	0.537	+/-0.176	0.516	0.748	0.582	$0.8470	1.1s	0.78	(!) brittle JSON (!) fails no-ad control
D	openai/gpt-oss-20b	0.525	+/-0.157	0.524	0.599	0.540	$0.0620	9.2s	0.80	(!) brittle JSON (!) fails no-ad control
D	deepseek/deepseek-r1-0528	0.520	+/-0.132	0.520	0.649	0.540	$1.1645	29.4s	0.84	(!) brittle JSON (!) fails no-ad control
D	xiaomi/mimo-v2.5-pro	0.496	+/-0.131	0.572	0.402	0.438	$0.4954	2.2s	0.90	(!) fails no-ad control
D	meta-llama/llama-3.3-70b-instruct	0.489	+/-0.153	0.493	0.521	0.493	$0.1107	1.3s	0.88	(!) brittle JSON
D	nvidia/nemotron-3-ultra-550b-a55b	0.485	+/-0.131	0.526	0.439	0.453	$0.7782	1.2s	0.89	(!) brittle JSON (!) fails no-ad control
D	meta-llama/llama-4-maverick	0.485	+/-0.165	0.466	0.628	0.522	$0.2206	1.1s	0.79	(!) brittle JSON (!) fails no-ad control
D	mistralai/mistral-large-2512	0.479	+/-0.157	0.447	0.854	0.547	$0.6184	3.3s	1.00	(!) fails no-ad control
E	stepfun/step-3.7-flash	0.473	+/-0.252	0.588	0.337	0.398	$0.4302	13.9s	0.56	(!) brittle JSON (!!) moderation blocked 15.2%
E	meta-llama/llama-4-scout	0.463	+/-0.141	0.491	0.439	0.442	$0.1111	0.9s	0.83	(!) brittle JSON (!) fails no-ad control
E	mistralai/codestral-2508	0.459	+/-0.125	0.438	0.679	0.504	$0.3585	0.9s	1.00	(!) fails no-ad control
E	qwen/qwen3-235b-a22b-2507	0.458	+/-0.118	0.439	0.626	0.499	$0.1078	2.5s	0.84	(!) brittle JSON (!) fails no-ad control
E	cohere/command-r-plus-08-2024	0.448	+/-0.162	0.534	0.334	0.381	$2.8505	1.1s	0.93
E	moonshotai/kimi-k2.6	0.448	+/-0.188	0.524	0.388	0.401	$1.8854	55.7s	0.63	(!) brittle JSON (!) fails no-ad control
E	qwen/qwen3.7-plus	0.447	+/-0.107	0.420	0.742	0.506	$0.7238	23.7s	0.80	(!) brittle JSON (!) fails no-ad control
E	xiaomi/mimo-v2.5	0.447	+/-0.149	0.445	0.598	0.471	$0.1692	4.2s	0.73	(!) brittle JSON (!) fails no-ad control
E	meituan/longcat-2.0	0.431	+/-0.154	0.483	0.361	0.390	$0.5573	10.5s	0.42	(!) brittle JSON
E	deepseek/deepseek-r1-distill-llama-70b	0.414	+/-0.165	0.454	0.371	0.387	$0.6913	27.2s	0.39	(!) brittle JSON (!) fails no-ad control
E	qwen/qwen3-14b	0.393	+/-0.171	0.406	0.394	0.385	$0.3989	14.4s	0.60	(!) brittle JSON (!) fails no-ad control
E	inclusionai/ring-2.6-1t	0.385	+/-0.198	0.447	0.299	0.337	$0.2468	10.8s	0.81	(!) brittle JSON
E	cohere/command-a	0.379	+/-0.128	0.357	0.611	0.425	$2.9415	1.7s	0.71	(!) brittle JSON (!) fails no-ad control
E	thinkingmachines/inkling-small	0.375	+/-0.153	0.402	0.366	0.358	$1.0333	18.5s	0.58	(!) brittle JSON (!) fails no-ad control
E	qwen/qwen3.8-max	0.340	+/-0.206	0.435	0.217	0.272	$4.2714	29.7s	0.73	(!) brittle JSON
E	microsoft/phi-4	0.337	+/-0.190	0.386	0.276	0.299	$0.0792	0.5s	0.98
E	thinkingmachines/inkling	0.294	+/-0.155	0.395	0.174	0.225	$3.2258	65.9s	0.35	(!) brittle JSON (!) fails no-ad control
E	openai/gpt-3.5-turbo	0.275	+/-0.140	0.260	0.484	0.310	$0.5564	1.4s	0.70	(!) brittle JSON (!) fails no-ad control
E	meta-llama/llama-3.1-8b-instruct	0.246	+/-0.118	0.243	0.325	0.259	$0.0572	0.9s	0.82	(!) brittle JSON (!) fails no-ad control
F	openai/o4-mini	0.189	+/-0.092	0.272	0.093	0.135	$2.1179	7.3s	0.07	(!) brittle JSON
G	mistralai/mistral-small-2603	0.020	+/-0.044	0.021	0.017	0.019	$0.1745	0.7s	1.00
G	qwen/qwen3-8b	0.000	+/-0.000	0.000	0.000	0.000	$0.3256	39.8s	0.10	(!) brittle JSON

Best Value (F0.5 per dollar)

Paid-tier only, ranked by F0.5 per dollar. Free-tier models are excluded here because F0.5 / 0 is undefined (none in this campaign’s roster came back at $0.00). No confidence tiers on this table, since a point ratio does not group cleanly, but the reliability flags still apply.

Rank	Model	F0.5/$	F0.5	F1	Cost / episode	Flags
1	openai/gpt-oss-120b	10.33	0.655	0.692	$0.0634	(!) brittle JSON (!) fails no-ad control
2	qwen/qwen3.7-flash	9.66	0.701	0.706	$0.0725
3	openai/gpt-oss-20b	8.46	0.525	0.540	$0.0620	(!) brittle JSON (!) fails no-ad control
4	google/gemma-4-26b-a4b-it	6.60	0.545	0.572	$0.0825	(!) brittle JSON (!) fails no-ad control
5	google/gemma-4-31b-it	5.60	0.662	0.667	$0.1181	(!) brittle JSON (!) fails no-ad control
6	openai/gpt-5.6-luna	5.25	0.736	0.754	$0.1402	(!) brittle JSON
7	google/gemini-2.5-flash-lite	5.00	0.566	0.623	$0.1132	(!) fails no-ad control
8	deepseek/deepseek-v4-flash-0731	4.43	0.656	0.601	$0.1482	(!) brittle JSON
9	meta-llama/llama-3.3-70b-instruct	4.42	0.489	0.493	$0.1107	(!) brittle JSON
10	meta-llama/llama-3.1-8b-instruct	4.30	0.246	0.259	$0.0572	(!) brittle JSON (!) fails no-ad control
11	microsoft/phi-4	4.26	0.337	0.299	$0.0792
12	qwen/qwen3-235b-a22b-2507	4.25	0.458	0.499	$0.1078	(!) brittle JSON (!) fails no-ad control
13	meta-llama/llama-4-scout	4.17	0.463	0.442	$0.1111	(!) brittle JSON (!) fails no-ad control
14	deepseek/deepseek-v4-flash	3.53	0.723	0.708	$0.2046	(!) brittle JSON
15	nvidia/nemotron-3-super-120b-a12b	2.72	0.597	0.577	$0.2194	(!) brittle JSON (!) fails no-ad control
16	xiaomi/mimo-v2.5	2.64	0.447	0.471	$0.1692	(!) brittle JSON (!) fails no-ad control
17	google/gemini-3.1-flash-lite	2.29	0.686	0.747	$0.2988	(!) fails no-ad control
18	deepseek/deepseek-v3.2	2.28	0.687	0.599	$0.3015
19	meta-llama/llama-4-maverick	2.20	0.485	0.522	$0.2206	(!) brittle JSON (!) fails no-ad control
20	google/gemini-3.5-flash-lite	2.12	0.774	0.776	$0.3643
21	minimax/minimax-m3	1.86	0.665	0.626	$0.3575	(!) brittle JSON (!) fails no-ad control
22	google/gemini-2.5-flash	1.69	0.641	0.694	$0.3796
23	inclusionai/ring-2.6-1t	1.56	0.385	0.337	$0.2468	(!) brittle JSON
24	tencent/hy3	1.41	0.548	0.489	$0.3876	(!) brittle JSON
25	mistralai/mistral-medium-3.1	1.38	0.669	0.699	$0.4847
26	qwen/qwen3.6-flash	1.36	0.738	0.747	$0.5435
27	mistralai/codestral-2508	1.28	0.459	0.504	$0.3585	(!) fails no-ad control
28	stepfun/step-3.7-flash	1.10	0.473	0.398	$0.4302	(!) brittle JSON (!!) moderation blocked 15.2%
29	deepseek/deepseek-v4-pro	1.06	0.708	0.680	$0.6692	(!) brittle JSON (!) fails no-ad control
30	xiaomi/mimo-v2.5-pro	1.00	0.496	0.438	$0.4954	(!) fails no-ad control
31	qwen/qwen3-14b	0.98	0.393	0.385	$0.3989	(!) brittle JSON (!) fails no-ad control
32	mistralai/mistral-large-2512	0.77	0.479	0.547	$0.6184	(!) fails no-ad control
33	meituan/longcat-2.0	0.77	0.431	0.390	$0.5573	(!) brittle JSON
34	z-ai/glm-5.2	0.77	0.568	0.606	$0.7376	(!) brittle JSON (!) fails no-ad control
35	claude-haiku-4-5-20251001	0.76	0.819	0.842	$1.0820
36	qwen/qwen3.5-plus-02-15	0.70	0.756	0.779	$1.0746
37	qwen/qwen3.5-27b	0.68	0.607	0.567	$0.8941	(!) brittle JSON
38	openai/gpt-5.6-terra	0.64	0.767	0.768	$1.1951	(!) brittle JSON (!) fails no-ad control
39	openai/gpt-5.4-mini	0.63	0.537	0.582	$0.8470	(!) brittle JSON (!) fails no-ad control
40	nvidia/nemotron-3-ultra-550b-a55b	0.62	0.485	0.453	$0.7782	(!) brittle JSON (!) fails no-ad control
41	qwen/qwen3.7-plus	0.62	0.447	0.506	$0.7238	(!) brittle JSON (!) fails no-ad control
42	deepseek/deepseek-r1-distill-llama-70b	0.60	0.414	0.387	$0.6913	(!) brittle JSON (!) fails no-ad control
43	qwen/qwen3.6-plus	0.54	0.622	0.607	$1.1534	(!) brittle JSON
44	openai/gpt-3.5-turbo	0.49	0.275	0.310	$0.5564	(!) brittle JSON (!) fails no-ad control
45	x-ai/grok-4.3	0.48	0.767	0.789	$1.6048
46	deepseek/deepseek-r1-0528	0.45	0.520	0.540	$1.1645	(!) brittle JSON (!) fails no-ad control
47	mistralai/mistral-medium-3-5	0.41	0.746	0.787	$1.8393
48	deepseek/deepseek-r1	0.37	0.552	0.574	$1.4743	(!) brittle JSON (!) fails no-ad control
49	thinkingmachines/inkling-small	0.36	0.375	0.358	$1.0333	(!) brittle JSON (!) fails no-ad control
50	claude-sonnet-5	0.33	0.708	0.732	$2.1452
51	meta/muse-spark-1.1	0.30	0.671	0.595	$2.2536
52	google/gemini-3.6-flash	0.27	0.788	0.796	$2.8863
53	qwen/qwen3.7-max	0.26	0.718	0.746	$2.8109
54	claude-sonnet-4-6	0.24	0.791	0.799	$3.2386
55	moonshotai/kimi-k2.6	0.24	0.448	0.401	$1.8854	(!) brittle JSON (!) fails no-ad control
56	x-ai/grok-4.5	0.23	0.781	0.797	$3.3554
57	google/gemini-3.5-flash	0.23	0.774	0.787	$3.4202
58	openai/gpt-5.4	0.21	0.592	0.634	$2.7907	(!) brittle JSON (!) fails no-ad control
59	openai/o3	0.17	0.563	0.457	$3.3100
60	google/gemini-2.5-pro	0.16	0.669	0.700	$4.1601	(!) fails no-ad control
61	cohere/command-r-plus-08-2024	0.16	0.448	0.381	$2.8505
62	google/gemini-3.1-pro-preview	0.15	0.732	0.752	$4.9160
63	claude-opus-4-8	0.14	0.728	0.760	$5.3661
64	claude-opus-5	0.13	0.725	0.748	$5.3761
65	claude-opus-4-7	0.13	0.704	0.731	$5.3293
66	cohere/command-a	0.13	0.379	0.425	$2.9415	(!) brittle JSON (!) fails no-ad control
67	moonshotai/kimi-k3	0.12	0.693	0.703	$5.9242	(!) brittle JSON (!) fails no-ad control
68	openai/gpt-5.6-sol	0.11	0.716	0.738	$6.2293	(!) brittle JSON (!) fails no-ad control
69	mistralai/mistral-small-2603	0.11	0.020	0.019	$0.1745
70	openai/gpt-5.5	0.10	0.780	0.784	$7.6842	(!) brittle JSON (!) fails no-ad control
71	thinkingmachines/inkling	0.09	0.294	0.225	$3.2258	(!) brittle JSON (!) fails no-ad control
72	openai/o4-mini	0.09	0.189	0.135	$2.1179	(!) brittle JSON
73	qwen/qwen3.8-max	0.08	0.340	0.272	$4.2714	(!) brittle JSON
74	claude-fable-5	0.07	0.752	0.783	$10.7552
75	qwen/qwen3-8b	0.00	0.000	0.000	$0.3256	(!) brittle JSON

Charts

Cost vs F1 (Pareto)

Each model is one colored point. Lower-left is unhelpful (expensive, inaccurate). Upper-left is the sweet spot (accurate, cheap). The legend below the chart shows each model’s color next to its F1 and cost-per-episode.

Source data: Best Accuracy, Best Value

Accuracy vs latency

F0.5 (y) against p50 latency (x, log scale). The cost Pareto above answers what accuracy costs in dollars; this one answers what it costs in wall-clock time. Upper-left is accurate and fast. MinusPod’s pipeline is offline, so a slow accurate model is usable, but the chart shows which models make you choose and which don’t. The OpenRouter latency caveat from the Metric Key applies.

Source data: Best Accuracy (F0.5), Latency tail (p50)

JSON schema compliance

Fraction of each model’s responses that parsed as a clean JSON array. 1.0 means every response came back exactly as requested; lower numbers mean the parser had to recover from markdown fences, object wrappers, or extra fields.

Source data: Per-Model Detail (JSON compliance field)

F1 by episode (heatmap)

F1 score for each (model, episode) pair. Greener is more accurate, redder is less. The no-ad episodes are excluded. They have no F1 because they’re PASS/FAIL negative controls.

Source data: Quick Comparison, Per-Episode Detail

Confidence calibration (heatmap)

One row per model, one column per self-reported confidence bin. Cell text is the actual hit rate at that bin plus the sample size; cell color is the calibration error (actual minus bin midpoint). Red cells mean the model claimed high confidence but was usually wrong; green is well-calibrated; blue is underconfident. Empty cells mean the model never produced a prediction in that bin. Models are sorted from most overconfident at the top to most underconfident at the bottom.

Source data: Confidence calibration table

Latency percentiles

p50, p90, p99, and max per model on a log scale. The gap between p99 and max indicates how heavy the tail is. For OpenRouter-routed models, the tail also includes upstream provider load.

Source data: Latency tail table

Cross-model agreement (window distribution)

Histogram of how many models flagged at least one ad per (episode, window). The left side is windows nobody flagged (clear non-ad content), the right side is windows everyone flagged (clear sponsor reads). Bars in the middle are contested (some models said yes, some said no) and are candidates for ensemble voting or manual review. This view is anonymous (bars don’t show which models contributed); the per-model breakdown is in the next chart.

Source data: Cross-model agreement table

Per-model alignment with majority

Stacked horizontal bar per model. Green + blue segments are windows where the model voted with the majority (true positives + true negatives); orange is windows where it voted yes but most others voted no (likely false positive / hallucination); red is windows where it voted no but most others voted yes (likely missed real ad). Right-edge label is alignment rate. High alignment means the model tracks consensus; low alignment is either insight or noise depending on whether those broken-from-consensus calls were right.

Source data: Per-model alignment with consensus table

Precision vs Recall (with F1 isocurves)

Scatter of precision (y) vs recall (x) for each model. Dashed gray lines are F1 isocurves; points on the same dashed line have the same F1. Top-right is ideal (high precision AND high recall). Top-left is cautious (high precision, low recall). Bottom-right is greedy (high recall, low precision). Useful for picking a model whose error profile matches your tolerance: precision-leaning for environments where false positives are expensive, recall-leaning for completeness-first.

Source data: Precision, recall, and FP/FN breakdown table

Boundary accuracy (start + end MAE)

Stacked horizontal bars per model: blue is mean absolute error on the predicted ad START in seconds, orange is the same for END. Total error labeled at the right. Sorted by total ascending so the cleanest boundaries are at the top. Skewed bars (start much larger than end, or vice versa) mean the model systematically overshoots on one side. Relevant if you cut audio downstream.

Source data: Boundary accuracy table

Token efficiency vs F1

Scatter of output tokens per detected ad (x, log scale) vs F1 (y). Upper-left is the efficient zone: high accuracy with few output tokens. Right-side points are reasoning-heavy models that emit chain-of-thought alongside their JSON. The chart answers whether the extra tokens buy more F1 or just burn output budget. A model that lands far right at modest F1 is paying for reasoning that didn’t help.

Source data: Output token efficiency table

Cost split (input vs output)

Stacked horizontal bars per model: blue is the input share of per-episode cost, orange is the output share, total labeled at the right, sorted by total ascending. Every model reads the same transcripts, so a long blue bar is an expensive input price and a long orange bar is a talkative model. Reasoning models show up as mostly orange.

Source data: Cost breakdown (input vs output) table

Trial variance (determinism check)

Horizontal bars of mean F1 stdev across episodes per model. All trials run at temperature 0.0 so well-behaved models cluster near zero. Bars are color-graded: green below 0.02 (effectively deterministic), yellow 0.02-0.05 (slight noise), red above 0.05 (single-trial F1 numbers from this model should be treated with suspicion). Dotted reference lines mark the 0.02 and 0.05 thresholds.

Source data: Trial variance (determinism check) table

Detection rate by ad length

Heatmap of model (row) vs ad-length bucket (column), cell = detection rate with sample size. Greener = caught more ads in that bucket; redder = missed more. Models are sorted by overall detection rate so the strongest are at the top. Empty (gray) cells mean that bucket had no truth ads for the corresponding model’s trials.

Source data: Detection rate by ad characteristic > By ad length table

Detection rate by ad position

Same shape as the ad-length heatmap, but columns are episode position (pre-roll / mid-roll / post-roll). A common pattern: pre-roll is easy because of clear show-intro transitions; post-roll is harder because models near the end of long episodes often produce shorter responses or run out of context to anchor on.

Source data: Detection rate by ad characteristic > By ad position table

Parser stress (extraction-method usage)

Heatmap of model (row) vs extraction-method (column), cell = number of responses parsed via that method. Columns are ordered by total usage. json_array_direct is the clean path; everything else is a recovery path the parser had to take because the model added markdown fences, wrapped the array in an object, or returned malformed JSON. Models near the top of the chart use the clean path most often. They are operationally easier to consume.

Source data: Parser stress test table

Failures and provider issues

139 call(s) failed out of 64125 total (0.22%). Failures are excluded from F1 / cost calculations, but they often surface real production-relevant gotchas worth knowing.

By category

Errors classified into coarse buckets so failure patterns are visible at a glance. A model showing up here doesn’t mean it’s broken. Some categories are provider-side (content moderation, rate limits) and tell you more about routing reliability than model quality.

Category	Calls	Affected models
Provider content moderation rejection	130	stepfun/step-3.7-flash
Rate-limited	7	thinkingmachines/inkling
Other	2	thinkingmachines/inkling

Per-model error count

Same errors grouped by model, with the failure rate as a fraction of that model’s total calls. Rates under 1% are usually one-off provider hiccups; rates above 5% suggest the model isn’t operationally viable for production with the current prompts and concurrency caps.

Model	Errors	of total
stepfun/step-3.7-flash	130	130/855 (15.2%)
thinkingmachines/inkling	9	9/855 (1.1%)

Sample messages (first 3 per category)

First three raw error messages per category, so you can see what the provider actually returned without grepping calls.jsonl. Messages are truncated to ~240 characters; full text lives in results/raw/calls.jsonl.

Provider content moderation rejection (130)

• stepfun/step-3.7-flash on ep-drink-champs-30c9a2d49f13 (trial 0, window 4): Error code: 451 - {‘error’: {‘message’: ‘Provider returned error’, ‘code’: 451, ‘metadata’: {‘raw’: ‘{“error”:{“message”:“The content you provided or machine outputted is blocked.”,“type”:“censorship_blocked”}}’, ‘provider_name’: ‘StepFun’,…
• stepfun/step-3.7-flash on ep-drink-champs-30c9a2d49f13 (trial 0, window 9): Error code: 451 - {‘error’: {‘message’: ‘Provider returned error’, ‘code’: 451, ‘metadata’: {‘raw’: ‘{“error”:{“message”:“The content you provided or machine outputted is blocked.”,“type”:“censorship_blocked”}}’, ‘provider_name’: ‘StepFun’,…
• stepfun/step-3.7-flash on ep-drink-champs-30c9a2d49f13 (trial 0, window 10): Error code: 451 - {‘error’: {‘message’: ‘Provider returned error’, ‘code’: 451, ‘metadata’: {‘raw’: ‘{“error”:{“message”:“The content you provided or machine outputted is blocked.”,“type”:“censorship_blocked”}}’, ‘provider_name’: ‘StepFun’,…
• … and 127 more

Rate-limited (7)

• thinkingmachines/inkling on ep-security-now-audio-2850b24903b2 (trial 2, window 2): Error code: 429 - {‘error’: {‘message’: ‘Provider returned error’, ‘code’: 429, ‘metadata’: {‘raw’: ’thinkingmachines/inkling is temporarily rate-limited upstream. Please retry shortly, or add your own key to accumulate your rate limits: ht…
• thinkingmachines/inkling on ep-security-now-audio-2850b24903b2 (trial 2, window 4): Error code: 429 - {‘error’: {‘message’: ‘Provider returned error’, ‘code’: 429, ‘metadata’: {‘raw’: ’thinkingmachines/inkling is temporarily rate-limited upstream. Please retry shortly, or add your own key to accumulate your rate limits: ht…
• thinkingmachines/inkling on ep-security-now-audio-2850b24903b2 (trial 2, window 9): Error code: 429 - {‘error’: {‘message’: ‘Provider returned error’, ‘code’: 429, ‘metadata’: {‘raw’: ’thinkingmachines/inkling is temporarily rate-limited upstream. Please retry shortly, or add your own key to accumulate your rate limits: ht…
• … and 4 more

Other (2)

• thinkingmachines/inkling on ep-drink-champs-30c9a2d49f13 (trial 2, window 36): Expecting value: line 241 column 1 (char 1320)
• thinkingmachines/inkling on ep-glt1412515089-373d5ba5007b (trial 4, window 7): Expecting value: line 271 column 1 (char 1485)

Errors resolved by retry

5032 work unit(s) errored at least once before succeeding (5541 errored attempts). These never reach the failure tables above because the retry’s successful row is what gets scored, but they are operationally real: a deployment without retry logic would have lost every one of them. Counts at or near a model’s full work-unit total usually mean a setup problem (auth, account gating, credit limits) that failed the whole first pass, not per-call flakiness.

Model	Work units	Errored attempts	Dominant cause
claude-opus-4-8	855	855	Auth failure (100%)
claude-opus-4-7	855	943	Auth failure (99%)
meta/muse-spark-1.1	855	855	Account gating (age confirmation) (100%)
deepseek/deepseek-r1-distill-llama-70b	855	856	Other (94%)
claude-sonnet-5	555	555	Auth failure (99%)
claude-sonnet-4-6	417	834	Auth failure (100%)
cohere/command-r-plus-08-2024	412	412	Credits exhausted (100%)
thinkingmachines/inkling	115	115	Rate-limited (86%)
deepseek/deepseek-r1	88	91	Credits exhausted (91%)
moonshotai/kimi-k2.6	4	4	Other (75%)
nvidia/nemotron-3-super-120b-a12b	4	4	Timeout (75%)
moonshotai/kimi-k3	3	3	Other (100%)
qwen/qwen3-8b	3	3	Other (67%)
deepseek/deepseek-r1-0528	3	3	Other (100%)
qwen/qwen3.6-plus	2	2	Other (100%)
stepfun/step-3.7-flash	2	2	Other (100%)
inclusionai/ring-2.6-1t	2	2	Other (100%)
qwen/qwen3.5-plus-02-15	1	1	Other (100%)
meituan/longcat-2.0	1	1	Other (100%)

Why this section exists

If you’re picking a model for production, an aggregate compliance score doesn’t tell you when the provider will simply refuse to answer. A few cases that have shown up here:

• Content moderation rejections (Alibaba on Qwen, Google on Gemma, sometimes others): the provider’s classifier blocks the prompt before the model runs. For ad detection on real podcast transcripts, this can happen on episodes with adult content, profanity, or politically sensitive topics. Rate is small but non-zero; plan for it.
• Deprecated parameters: the Claude 4.x family rejects temperature. The benchmark memoizes this per-process and retries without, but it tells you which models you cannot pass legacy sampling controls to.
• Rate limits: tail-latency or 429s under load. Not a model-quality issue, but determines whether a given provider is operationally viable for your throughput.

Precision, recall, and FP/FN breakdown

F1 collapses two failure modes into one number. A precision-leaning model misses ads but rarely flags non-ads; a recall-leaning model catches everything at the cost of false positives. Production tradeoffs hinge on which one you can tolerate.

Column key

Column	Meaning	Range
TP (true positive)	Predicted an ad and a real ad existed at that span (IoU >= 0.5)	0 to total truth ads
FP (false positive)	Predicted an ad where no real ad existed	0 to total predictions
FN (false negative)	Missed a real ad entirely (no prediction matched it at IoU >= 0.5)	0 to total truth ads
Precision	TP / (TP + FP). Of the ads the model claimed, how many were real? Higher means fewer false positives.	0.000 to 1.000
Recall	TP / (TP + FN). Of the real ads, how many did the model find? Higher means fewer misses.	0.000 to 1.000

Reading the table: high precision + low recall means the model is cautious. It rarely flags something that isn’t an ad, but misses real ads. High recall + low precision means the opposite: catches everything but invents false positives. F1 is the harmonic mean of the two and rewards models that do both well.

Model	Precision	Recall	TP	FP	FN
claude-haiku-4-5-20251001	0.807	0.899	223	62	32
claude-sonnet-4-6	0.795	0.846	211	58	44
x-ai/grok-4.5	0.777	0.853	215	65	40
google/gemini-3.6-flash	0.788	0.840	211	60	44
x-ai/grok-4.3	0.755	0.847	213	72	42
google/gemini-3.5-flash	0.771	0.838	211	66	44
mistralai/mistral-medium-3-5	0.724	0.886	221	94	34
openai/gpt-5.5	0.783	0.819	203	51	52
claude-fable-5	0.740	0.878	218	94	37
qwen/qwen3.5-plus-02-15	0.745	0.844	214	72	41
google/gemini-3.5-flash-lite	0.786	0.813	205	65	50
openai/gpt-5.6-terra	0.772	0.808	199	62	56
claude-opus-4-8	0.715	0.857	211	86	44
openai/gpt-5.6-luna	0.730	0.824	203	69	52
google/gemini-3.1-pro-preview	0.726	0.826	207	80	48
claude-opus-5	0.718	0.845	215	84	40
google/gemini-3.1-flash-lite	0.653	0.918	229	136	26
qwen/qwen3.6-flash	0.737	0.788	190	72	65
qwen/qwen3.7-max	0.706	0.826	203	97	52
openai/gpt-5.6-sol	0.709	0.816	205	92	50
claude-sonnet-5	0.698	0.812	203	100	52
claude-opus-4-7	0.693	0.816	208	96	47
deepseek/deepseek-v4-flash	0.746	0.717	175	63	80
qwen/qwen3.7-flash	0.702	0.740	181	78	74
moonshotai/kimi-k3	0.694	0.756	180	81	75
google/gemini-2.5-pro	0.656	0.794	198	112	57
mistralai/mistral-medium-3.1	0.655	0.790	195	114	60
google/gemini-2.5-flash	0.612	0.859	220	150	35
openai/gpt-oss-120b	0.636	0.807	191	159	64
deepseek/deepseek-v4-pro	0.739	0.665	162	58	93
google/gemma-4-31b-it	0.665	0.701	178	101	77
openai/gpt-5.4	0.574	0.776	199	155	56
minimax/minimax-m3	0.720	0.612	155	71	100
google/gemini-2.5-flash-lite	0.540	0.854	207	240	48
qwen/qwen3.6-plus	0.645	0.619	159	85	96
z-ai/glm-5.2	0.555	0.769	189	184	66
deepseek/deepseek-v4-flash-0731	0.720	0.557	128	47	127
deepseek/deepseek-v3.2	0.792	0.552	139	46	116
meta/muse-spark-1.1	0.765	0.530	137	41	118
openai/gpt-5.4-mini	0.516	0.748	191	262	64
nvidia/nemotron-3-super-120b-a12b	0.624	0.577	132	74	123
deepseek/deepseek-r1	0.546	0.674	171	196	84
google/gemma-4-26b-a4b-it	0.532	0.651	155	143	100
qwen/qwen3.5-27b	0.675	0.563	135	74	120
mistralai/mistral-large-2512	0.447	0.854	217	426	38
deepseek/deepseek-r1-0528	0.520	0.649	162	203	93
openai/gpt-oss-20b	0.524	0.599	137	149	118
meta-llama/llama-4-maverick	0.466	0.628	156	189	99
qwen/qwen3.7-plus	0.420	0.742	186	329	69
mistralai/codestral-2508	0.438	0.679	163	250	92
qwen/qwen3-235b-a22b-2507	0.439	0.626	150	226	105
meta-llama/llama-3.3-70b-instruct	0.493	0.521	120	152	135
tencent/hy3	0.615	0.441	103	40	152
xiaomi/mimo-v2.5	0.445	0.598	162	250	93
openai/o3	0.710	0.371	94	22	161
nvidia/nemotron-3-ultra-550b-a55b	0.526	0.439	102	99	153
meta-llama/llama-4-scout	0.491	0.439	105	168	150
xiaomi/mimo-v2.5-pro	0.572	0.402	103	80	152
cohere/command-a	0.357	0.611	134	351	121
moonshotai/kimi-k2.6	0.524	0.388	90	80	165
stepfun/step-3.7-flash	0.588	0.337	59	9	136
meituan/longcat-2.0	0.483	0.361	91	80	164
deepseek/deepseek-r1-distill-llama-70b	0.454	0.371	95	101	160
qwen/qwen3-14b	0.406	0.394	90	165	165
cohere/command-r-plus-08-2024	0.534	0.334	88	59	167
thinkingmachines/inkling-small	0.402	0.366	85	108	170
inclusionai/ring-2.6-1t	0.447	0.299	64	82	191
openai/gpt-3.5-turbo	0.260	0.484	112	489	143
microsoft/phi-4	0.386	0.276	56	103	199
qwen/qwen3.8-max	0.435	0.217	48	16	207
meta-llama/llama-3.1-8b-instruct	0.243	0.325	65	347	190
thinkingmachines/inkling	0.395	0.174	35	26	195
openai/o4-mini	0.272	0.093	20	30	235
mistralai/mistral-small-2603	0.021	0.017	5	15	250
qwen/qwen3-8b	0.000	0.000	0	1	255

Boundary accuracy

For ads that match the truth at IoU >= 0.5, how far off were the predicted start and end timestamps? Lower is better. A model can hit F1 cleanly while still being 20s off on every boundary. Bad for any pipeline that cuts the audio.

MAE is size of the miss; bias is its direction (mean of predicted minus truth). A negative start bias or positive end bias means the cut extends past the ad and eats surrounding content; the opposite signs mean ad audio is left in. MinusPod cuts what the model flags, so a model whose bias points outward over-cuts even when its MAE looks acceptable. Bias near zero with a large MAE means the misses are random rather than systematic.

Model	Start MAE (s)	End MAE (s)	Start bias (s)	End bias (s)
mistralai/mistral-small-2603	0.00	0.39	+0.00	-0.39
stepfun/step-3.7-flash	0.82	1.20	-0.08	+0.10
qwen/qwen3-14b	1.59	4.08	+1.45	-2.90
google/gemini-3.1-flash-lite	2.87	3.39	+0.27	+0.29
claude-haiku-4-5-20251001	3.53	3.59	+2.00	+0.54
google/gemini-3.5-flash-lite	3.03	4.19	+2.41	+0.42
qwen/qwen3.6-plus	5.37	2.29	-4.33	-1.31
google/gemini-2.5-flash	4.28	3.50	+0.70	+0.26
qwen/qwen3.7-max	4.23	3.81	+0.12	-0.27
qwen/qwen3.7-flash	5.73	2.53	-1.32	+1.55
google/gemini-3.5-flash	5.05	3.26	-1.77	+1.55
mistralai/mistral-medium-3-5	4.18	4.24	+1.01	-2.55
google/gemini-3.6-flash	5.18	3.29	-0.42	+1.51
deepseek/deepseek-r1	5.14	3.62	+1.47	-0.14
x-ai/grok-4.5	5.80	3.13	-2.06	+1.37
tencent/hy3	5.51	3.49	-1.52	+3.07
thinkingmachines/inkling	6.39	2.69	-2.98	-0.12
meta-llama/llama-3.3-70b-instruct	3.92	5.24	+3.90	-4.01
deepseek/deepseek-v4-flash	4.62	4.54	-0.82	-0.43
mistralai/mistral-large-2512	4.01	5.18	+0.70	-0.72
x-ai/grok-4.3	5.30	3.98	-1.09	-0.17
deepseek/deepseek-r1-0528	5.69	3.81	+0.96	+0.18
qwen/qwen3.6-flash	7.35	2.20	-2.70	-0.43
deepseek/deepseek-v4-pro	5.25	4.32	-1.25	-0.36
openai/gpt-5.6-terra	5.91	3.70	-2.55	+0.91
qwen/qwen3.5-plus-02-15	7.39	2.24	-3.71	-0.51
claude-sonnet-4-6	6.62	3.09	-2.67	+1.31
minimax/minimax-m3	6.94	2.78	+2.60	-0.96
claude-fable-5	6.27	3.50	-2.14	+2.37
claude-opus-4-8	7.29	2.64	-3.39	+0.35
mistralai/mistral-medium-3.1	4.33	5.78	+0.29	-1.21
claude-sonnet-5	6.70	3.51	-1.79	+1.25
deepseek/deepseek-v4-flash-0731	4.40	5.89	-2.68	-1.76
openai/gpt-5.5	7.15	3.40	-4.54	+1.05
moonshotai/kimi-k2.6	8.57	1.99	-3.65	+0.19
google/gemini-3.1-pro-preview	5.26	5.38	-2.34	+3.56
qwen/qwen3.5-27b	7.72	3.09	-2.50	+0.42
qwen/qwen3.7-plus	7.18	3.67	-4.32	-3.14
deepseek/deepseek-v3.2	7.26	3.74	-2.56	-1.07
thinkingmachines/inkling-small	4.58	6.57	-0.60	-3.53
qwen/qwen3.8-max	3.85	7.32	-3.00	+2.71
openai/gpt-5.6-luna	7.70	3.79	-4.95	+1.30
moonshotai/kimi-k3	6.07	5.46	-2.32	+0.70
z-ai/glm-5.2	6.21	5.51	-2.41	-0.89
nvidia/nemotron-3-super-120b-a12b	5.70	6.05	-0.21	+2.80
claude-opus-4-7	9.16	2.63	-5.32	+0.16
meituan/longcat-2.0	4.24	7.59	+0.16	-2.15
openai/gpt-5.6-sol	8.03	3.84	-3.99	+0.78
meta/muse-spark-1.1	7.97	4.25	-2.36	-0.18
google/gemini-2.5-pro	9.05	3.35	-6.00	+1.56
google/gemini-2.5-flash-lite	5.88	6.54	-0.80	-2.00
google/gemma-4-26b-a4b-it	1.93	10.74	+0.85	-6.04
claude-opus-5	11.05	2.25	-5.47	+1.15
nvidia/nemotron-3-ultra-550b-a55b	6.53	6.95	-0.54	-5.78
deepseek/deepseek-r1-distill-llama-70b	5.47	8.22	+4.33	-4.65
openai/gpt-5.4	7.40	6.35	-3.59	-2.92
xiaomi/mimo-v2.5	6.57	7.26	-2.51	-1.46
cohere/command-a	5.47	8.42	-3.14	-6.90
google/gemma-4-31b-it	9.18	4.80	-7.51	-2.55
openai/gpt-oss-120b	8.98	5.07	+1.17	-1.16
qwen/qwen3-235b-a22b-2507	4.31	10.26	-0.90	-5.13
xiaomi/mimo-v2.5-pro	6.98	8.43	-1.50	-1.89
openai/gpt-oss-20b	8.11	7.34	+1.01	+0.14
mistralai/codestral-2508	5.37	10.76	-0.20	-6.10
openai/o3	11.31	4.95	-7.66	+0.30
openai/gpt-5.4-mini	6.87	11.19	-4.19	-7.32
inclusionai/ring-2.6-1t	2.47	15.62	+0.45	-9.13
meta-llama/llama-4-maverick	5.97	13.04	-2.84	-5.87
microsoft/phi-4	7.04	12.22	+6.53	+1.67
openai/o4-mini	5.95	13.31	-1.44	-7.90
openai/gpt-3.5-turbo	4.94	15.47	+4.78	-11.64
meta-llama/llama-3.1-8b-instruct	14.50	6.49	-0.38	-2.19
cohere/command-r-plus-08-2024	6.02	15.86	-0.13	-7.48
meta-llama/llama-4-scout	6.61	17.17	+3.07	-13.87

Confidence calibration

Models include a self-reported confidence on each detected ad. A well-calibrated model should be right ~95% of the time when it claims 0.95 confidence. The table below bins each model’s predictions and shows the actual hit rate (fraction that were true positives at IoU >= 0.5). A bin near 1.0 is well-calibrated; a low number with a high count means the model is overconfident.

Model	0.00-0.70	0.70-0.90	0.90-0.95	0.95-0.99	0.99+	total
claude-fable-5	0.00 (n=17)	0.00 (n=45)	0.62 (n=8)	0.88 (n=242)	–	312
claude-haiku-4-5-20251001	0.00 (n=1)	0.36 (n=11)	0.25 (n=12)	0.82 (n=228)	0.88 (n=33)	285
claude-opus-4-7	0.00 (n=8)	0.00 (n=17)	0.00 (n=31)	0.84 (n=239)	0.89 (n=9)	304
claude-opus-4-8	0.00 (n=9)	0.00 (n=25)	0.17 (n=23)	0.86 (n=240)	–	297
claude-opus-5	0.00 (n=7)	0.00 (n=36)	0.00 (n=15)	0.89 (n=241)	–	299
claude-sonnet-4-6	0.00 (n=2)	0.04 (n=25)	0.83 (n=6)	0.88 (n=198)	0.79 (n=38)	269
claude-sonnet-5	0.00 (n=18)	0.05 (n=39)	0.29 (n=14)	0.85 (n=232)	–	303
cohere/command-a	0.00 (n=63)	0.00 (n=30)	0.00 (n=45)	0.35 (n=361)	0.45 (n=20)	519
cohere/command-r-plus-08-2024	–	–	0.67 (n=15)	0.68 (n=56)	0.53 (n=76)	147
deepseek/deepseek-r1	–	0.10 (n=10)	0.00 (n=20)	0.28 (n=185)	0.74 (n=160)	375
deepseek/deepseek-r1-0528	0.00 (n=8)	0.00 (n=12)	0.00 (n=21)	0.26 (n=193)	0.77 (n=145)	379
deepseek/deepseek-r1-distill-llama-70b	–	0.40 (n=10)	0.00 (n=4)	0.48 (n=151)	0.53 (n=36)	201
deepseek/deepseek-v3.2	–	–	0.00 (n=6)	0.73 (n=71)	0.81 (n=108)	185
deepseek/deepseek-v4-flash	0.00 (n=9)	0.00 (n=6)	0.25 (n=4)	0.78 (n=182)	0.86 (n=37)	238
deepseek/deepseek-v4-flash-0731	0.33 (n=3)	0.00 (n=5)	–	0.76 (n=158)	0.78 (n=9)	175
deepseek/deepseek-v4-pro	0.00 (n=11)	0.00 (n=9)	0.00 (n=3)	0.80 (n=164)	0.88 (n=34)	221
google/gemini-2.5-flash	–	0.00 (n=20)	0.15 (n=93)	0.81 (n=162)	0.79 (n=95)	370
google/gemini-2.5-flash-lite	–	0.00 (n=2)	0.00 (n=60)	0.53 (n=391)	–	453
google/gemini-2.5-pro	–	0.00 (n=10)	0.00 (n=24)	0.54 (n=69)	0.75 (n=215)	318
google/gemini-3.1-flash-lite	–	0.00 (n=3)	0.00 (n=6)	0.30 (n=122)	0.81 (n=239)	370
google/gemini-3.1-pro-preview	–	0.00 (n=13)	0.00 (n=6)	0.18 (n=33)	0.86 (n=235)	287
google/gemini-3.5-flash	–	0.00 (n=7)	0.00 (n=1)	0.50 (n=50)	0.85 (n=219)	277
google/gemini-3.5-flash-lite	–	0.00 (n=1)	0.00 (n=2)	0.77 (n=230)	0.76 (n=37)	270
google/gemini-3.6-flash	–	0.00 (n=4)	0.00 (n=1)	0.74 (n=207)	0.98 (n=59)	271
google/gemma-4-26b-a4b-it	–	0.00 (n=9)	0.12 (n=41)	0.56 (n=142)	0.63 (n=111)	303
google/gemma-4-31b-it	–	0.00 (n=7)	0.07 (n=27)	0.54 (n=84)	0.79 (n=166)	284
inclusionai/ring-2.6-1t	0.00 (n=4)	0.06 (n=16)	0.25 (n=16)	0.56 (n=103)	0.14 (n=7)	146
meituan/longcat-2.0	0.00 (n=2)	0.00 (n=13)	0.08 (n=13)	0.63 (n=143)	–	171
meta-llama/llama-3.1-8b-instruct	0.00 (n=17)	0.22 (n=9)	0.00 (n=23)	0.17 (n=361)	0.00 (n=8)	418
meta-llama/llama-3.3-70b-instruct	–	0.00 (n=12)	0.00 (n=21)	0.46 (n=157)	0.59 (n=82)	272
meta-llama/llama-4-maverick	0.00 (n=21)	0.00 (n=65)	0.00 (n=27)	0.66 (n=237)	–	350
meta-llama/llama-4-scout	0.00 (n=13)	0.00 (n=1)	0.09 (n=23)	0.43 (n=184)	0.45 (n=53)	274
meta/muse-spark-1.1	–	0.00 (n=9)	0.11 (n=9)	0.86 (n=159)	0.00 (n=1)	178
microsoft/phi-4	–	–	0.00 (n=6)	0.37 (n=153)	–	159
minimax/minimax-m3	0.00 (n=16)	0.05 (n=22)	0.17 (n=6)	0.80 (n=174)	0.93 (n=14)	232
mistralai/codestral-2508	–	0.00 (n=2)	0.09 (n=205)	0.69 (n=211)	–	418
mistralai/mistral-large-2512	0.00 (n=15)	0.00 (n=86)	0.00 (n=59)	0.20 (n=290)	0.80 (n=198)	648
mistralai/mistral-medium-3-5	–	–	0.00 (n=4)	0.53 (n=85)	0.78 (n=226)	315
mistralai/mistral-medium-3.1	–	0.00 (n=5)	0.06 (n=17)	0.68 (n=275)	0.67 (n=12)	309
mistralai/mistral-small-2603	–	–	–	0.00 (n=5)	0.33 (n=15)	20
moonshotai/kimi-k2.6	0.00 (n=25)	0.00 (n=18)	0.00 (n=4)	0.49 (n=79)	0.93 (n=55)	181
moonshotai/kimi-k3	0.00 (n=5)	0.03 (n=29)	0.00 (n=17)	0.84 (n=204)	1.00 (n=7)	262
nvidia/nemotron-3-super-120b-a12b	0.00 (n=1)	0.00 (n=4)	0.14 (n=7)	0.60 (n=151)	0.89 (n=45)	208
nvidia/nemotron-3-ultra-550b-a55b	0.00 (n=18)	0.00 (n=4)	–	0.50 (n=171)	0.84 (n=19)	212
openai/gpt-3.5-turbo	–	–	0.00 (n=31)	0.20 (n=553)	0.00 (n=76)	660
openai/gpt-5.4	0.00 (n=58)	0.03 (n=40)	0.06 (n=16)	0.55 (n=47)	0.83 (n=206)	367
openai/gpt-5.4-mini	0.00 (n=90)	0.00 (n=66)	0.00 (n=36)	0.43 (n=53)	0.76 (n=220)	465
openai/gpt-5.5	0.00 (n=7)	0.09 (n=22)	0.57 (n=7)	0.86 (n=84)	0.92 (n=136)	256
openai/gpt-5.6-luna	0.00 (n=7)	0.17 (n=23)	0.00 (n=7)	0.36 (n=25)	0.90 (n=210)	272
openai/gpt-5.6-sol	0.00 (n=10)	0.14 (n=22)	0.00 (n=12)	0.06 (n=32)	0.87 (n=230)	306
openai/gpt-5.6-terra	0.00 (n=7)	0.00 (n=19)	0.20 (n=10)	0.44 (n=18)	0.89 (n=212)	266
openai/gpt-oss-120b	0.00 (n=1)	0.00 (n=4)	0.00 (n=3)	0.34 (n=146)	0.70 (n=202)	356
openai/gpt-oss-20b	–	0.00 (n=2)	0.07 (n=15)	0.36 (n=135)	0.62 (n=140)	292
openai/o3	0.00 (n=2)	0.00 (n=4)	0.69 (n=16)	0.88 (n=91)	1.00 (n=3)	116
openai/o4-mini	0.00 (n=1)	0.33 (n=3)	0.00 (n=6)	0.49 (n=39)	0.00 (n=1)	50
qwen/qwen3-14b	0.00 (n=16)	0.00 (n=18)	0.60 (n=5)	0.39 (n=221)	–	260
qwen/qwen3-235b-a22b-2507	0.00 (n=19)	0.00 (n=3)	0.00 (n=2)	0.42 (n=351)	0.17 (n=23)	398
qwen/qwen3-8b	–	–	–	0.00 (n=1)	–	1
qwen/qwen3.5-27b	–	0.20 (n=5)	0.12 (n=16)	0.70 (n=188)	–	209
qwen/qwen3.5-plus-02-15	–	0.00 (n=9)	0.00 (n=4)	0.79 (n=258)	0.67 (n=15)	286
qwen/qwen3.6-flash	0.00 (n=1)	0.00 (n=7)	0.00 (n=6)	0.77 (n=245)	0.33 (n=3)	262
qwen/qwen3.6-plus	–	–	1.00 (n=1)	0.66 (n=229)	0.57 (n=14)	244
qwen/qwen3.7-flash	–	0.00 (n=9)	0.36 (n=11)	0.75 (n=237)	0.00 (n=2)	259
qwen/qwen3.7-max	0.00 (n=9)	0.03 (n=29)	0.29 (n=7)	0.78 (n=255)	–	300
qwen/qwen3.7-plus	0.00 (n=184)	0.12 (n=26)	0.00 (n=2)	0.54 (n=342)	–	554
qwen/qwen3.8-max	–	–	–	0.71 (n=48)	0.88 (n=16)	64
stepfun/step-3.7-flash	–	–	0.80 (n=5)	0.87 (n=62)	1.00 (n=1)	68
tencent/hy3	0.11 (n=9)	0.00 (n=6)	0.00 (n=2)	0.81 (n=125)	1.00 (n=1)	143
thinkingmachines/inkling	0.00 (n=1)	0.17 (n=6)	0.00 (n=4)	0.48 (n=25)	0.85 (n=26)	62
thinkingmachines/inkling-small	0.00 (n=38)	0.00 (n=9)	0.00 (n=15)	0.65 (n=129)	0.05 (n=19)	210
x-ai/grok-4.3	–	0.25 (n=4)	0.07 (n=14)	0.79 (n=246)	0.81 (n=21)	285
x-ai/grok-4.5	–	0.00 (n=7)	0.00 (n=12)	0.84 (n=227)	0.74 (n=34)	280
xiaomi/mimo-v2.5	0.00 (n=26)	0.02 (n=44)	0.00 (n=30)	0.42 (n=258)	0.65 (n=80)	438
xiaomi/mimo-v2.5-pro	0.00 (n=1)	0.00 (n=16)	0.00 (n=11)	0.63 (n=119)	0.76 (n=37)	184
z-ai/glm-5.2	0.03 (n=113)	0.00 (n=29)	0.00 (n=8)	0.76 (n=220)	0.78 (n=23)	393

See report_assets/calibration.svg for the visual reliability diagram.

Latency tail

Median latency hides outliers. p99 and max are what determines queue depth and worst-case user wait. For OpenRouter-routed models the tail also reflects upstream provider load, not just model compute.

Model	p50	p90	p95	p99	max
microsoft/phi-4	0.49s	3.14s	5.50s	69.96s	101.52s
mistralai/mistral-small-2603	0.66s	0.98s	1.16s	2.27s	6.56s
google/gemini-3.5-flash-lite	0.67s	1.23s	1.42s	1.77s	2.60s
mistralai/mistral-medium-3.1	0.74s	5.83s	7.65s	10.73s	16.46s
google/gemini-2.5-flash	0.83s	1.91s	2.21s	2.98s	15.68s
google/gemini-2.5-flash-lite	0.84s	1.50s	1.75s	3.01s	6.73s
google/gemini-3.1-flash-lite	0.85s	1.57s	2.11s	12.92s	46.86s
mistralai/codestral-2508	0.88s	2.14s	2.54s	3.45s	8.33s
meta-llama/llama-4-scout	0.93s	3.35s	4.49s	7.51s	49.85s
meta-llama/llama-3.1-8b-instruct	0.93s	2.92s	4.34s	5.14s	9.27s
mistralai/mistral-medium-3-5	1.10s	2.80s	3.29s	4.54s	14.30s
meta-llama/llama-4-maverick	1.12s	2.72s	3.59s	4.60s	37.92s
openai/gpt-5.4-mini	1.13s	1.86s	2.45s	3.16s	6.40s
cohere/command-r-plus-08-2024	1.13s	13.03s	16.24s	24.51s	62.18s
nvidia/nemotron-3-ultra-550b-a55b	1.24s	3.00s	5.94s	14.87s	68.31s
meta-llama/llama-3.3-70b-instruct	1.29s	5.55s	8.60s	19.20s	41.61s
openai/gpt-3.5-turbo	1.38s	1.82s	1.96s	2.31s	3.06s
google/gemma-4-26b-a4b-it	1.40s	4.17s	6.52s	13.03s	99.69s
openai/gpt-5.4	1.43s	2.03s	2.57s	3.84s	15.05s
minimax/minimax-m3	1.68s	4.84s	6.67s	15.92s	146.60s
cohere/command-a	1.72s	2.54s	2.92s	4.17s	5.73s
deepseek/deepseek-v3.2	1.74s	8.05s	10.84s	15.95s	88.47s
openai/gpt-5.6-terra	2.10s	5.19s	6.71s	9.29s	27.15s
xiaomi/mimo-v2.5-pro	2.19s	4.89s	6.41s	11.76s	55.00s
google/gemma-4-31b-it	2.45s	7.79s	11.72s	24.30s	164.44s
qwen/qwen3-235b-a22b-2507	2.51s	5.52s	7.43s	15.66s	18.61s
mistralai/mistral-large-2512	3.28s	6.54s	7.58s	11.79s	35.55s
openai/gpt-5.6-luna	3.46s	8.81s	11.35s	20.00s	83.66s
z-ai/glm-5.2	3.55s	15.39s	24.22s	50.27s	202.19s
claude-opus-5	3.67s	63.46s	64.55s	182.66s	244.12s
openai/gpt-5.6-sol	3.72s	9.39s	11.86s	20.99s	76.93s
x-ai/grok-4.3	3.73s	8.76s	10.34s	12.67s	20.15s
claude-sonnet-4-6	4.11s	63.07s	83.49s	170.46s	351.49s
xiaomi/mimo-v2.5	4.16s	12.88s	21.43s	62.00s	201.00s
meta/muse-spark-1.1	4.31s	10.89s	13.59s	17.76s	20.43s
google/gemini-3.6-flash	4.51s	8.72s	10.24s	14.59s	19.88s
claude-opus-4-7	4.60s	64.27s	66.02s	183.69s	245.11s
openai/gpt-5.5	5.36s	16.34s	23.68s	45.89s	61.80s
google/gemini-3.5-flash	5.59s	9.27s	11.20s	16.40s	20.77s
deepseek/deepseek-v4-flash	6.27s	32.38s	40.26s	88.86s	224.85s
openai/gpt-oss-120b	6.54s	30.43s	44.55s	78.52s	1677.95s
claude-fable-5	6.61s	64.22s	68.22s	183.43s	244.88s
openai/o3	6.66s	17.18s	20.44s	29.16s	176.74s
openai/o4-mini	7.34s	20.82s	25.55s	54.13s	73.95s
qwen/qwen3.6-flash	7.55s	17.78s	19.48s	24.46s	35.86s
claude-opus-4-8	7.76s	63.15s	68.92s	104.92s	245.48s
google/gemini-3.1-pro-preview	8.35s	16.39s	20.73s	28.85s	32.09s
claude-sonnet-5	8.48s	62.69s	66.02s	146.69s	248.07s
openai/gpt-oss-20b	9.18s	38.09s	51.77s	72.93s	571.38s
qwen/qwen3.7-flash	10.14s	24.59s	27.58s	31.94s	38.97s
meituan/longcat-2.0	10.53s	69.38s	77.08s	88.11s	122.09s
inclusionai/ring-2.6-1t	10.75s	32.71s	35.76s	41.10s	137.68s
x-ai/grok-4.5	10.86s	42.26s	52.43s	83.77s	120.19s
deepseek/deepseek-v4-flash-0731	13.20s	44.63s	61.85s	130.27s	326.89s
moonshotai/kimi-k3	13.78s	55.72s	96.34s	160.79s	298.31s
stepfun/step-3.7-flash	13.87s	33.94s	35.00s	36.93s	53.50s
google/gemini-2.5-pro	14.22s	27.92s	34.09s	130.41s	154.37s
qwen/qwen3-14b	14.39s	42.03s	59.03s	163.41s	246.83s
deepseek/deepseek-v4-pro	16.17s	56.57s	68.15s	95.60s	165.32s
thinkingmachines/inkling-small	18.51s	28.02s	30.60s	36.06s	82.20s
tencent/hy3	22.27s	55.36s	59.98s	88.55s	126.83s
nvidia/nemotron-3-super-120b-a12b	22.33s	182.06s	248.36s	416.82s	1201.50s
qwen/qwen3.7-max	23.41s	54.27s	62.65s	75.47s	76.89s
qwen/qwen3.7-plus	23.68s	52.61s	62.94s	71.06s	72.60s
claude-haiku-4-5-20251001	24.23s	105.59s	135.48s	248.50s	680.52s
deepseek/deepseek-r1-distill-llama-70b	27.21s	68.34s	97.56s	138.12s	158.13s
qwen/qwen3.5-plus-02-15	29.18s	46.60s	52.05s	60.74s	94.18s
deepseek/deepseek-r1-0528	29.36s	172.78s	208.22s	296.84s	415.53s
qwen/qwen3.8-max	29.66s	87.68s	90.22s	93.28s	97.65s
deepseek/deepseek-r1	35.66s	170.50s	195.41s	276.94s	428.59s
qwen/qwen3.6-plus	36.91s	70.52s	70.77s	71.10s	71.67s
qwen/qwen3.5-27b	37.62s	106.00s	126.80s	163.08s	989.60s
qwen/qwen3-8b	39.80s	133.65s	170.51s	252.88s	1187.74s
moonshotai/kimi-k2.6	55.69s	126.30s	164.28s	249.82s	485.45s
thinkingmachines/inkling	65.91s	227.48s	324.88s	508.18s	1063.70s

Output token efficiency

How many output tokens the model spent per detected ad. Lower is more concise (the model finds an ad and returns the JSON). Higher means the model is producing a lot of text the parser will discard, which costs you whether or not the answer is right.

Model	Total output tokens	Ads detected	Tokens / ad	Cost / TP
claude-opus-4-7	29,485	615	48	$0.0256
claude-sonnet-5	36,216	612	59	$0.0106
meta-llama/llama-3.3-70b-instruct	28,052	470	60	$0.0009
google/gemini-3.1-flash-lite	50,395	842	60	$0.0013
claude-opus-4-8	36,858	605	61	$0.0254
claude-haiku-4-5-20251001	45,648	716	64	$0.0049
claude-fable-5	39,151	566	69	$0.0493
mistralai/codestral-2508	57,458	820	70	$0.0022
google/gemini-3.5-flash-lite	47,096	665	71	$0.0018
claude-opus-5	38,844	547	71	$0.0250
mistralai/mistral-medium-3.1	51,122	705	73	$0.0025
openai/gpt-3.5-turbo	56,074	767	73	$0.0050
deepseek/deepseek-v3.2	31,971	432	74	$0.0022
google/gemma-4-26b-a4b-it	37,850	495	76	$0.0005
cohere/command-r-plus-08-2024	15,352	196	78	$0.0324
mistralai/mistral-medium-3-5	65,551	833	79	$0.0083
meta-llama/llama-4-maverick	38,620	476	81	$0.0014
google/gemini-2.5-flash	80,965	975	83	$0.0017
claude-sonnet-4-6	43,175	516	84	$0.0153
cohere/command-a	60,841	655	93	$0.0220
openai/gpt-5.4-mini	56,545	599	94	$0.0044
google/gemma-4-31b-it	54,323	569	95	$0.0007
openai/gpt-5.4	45,697	470	97	$0.0140
mistralai/mistral-large-2512	126,901	1280	99	$0.0028
qwen/qwen3-235b-a22b-2507	65,771	633	104	$0.0007
google/gemini-2.5-flash-lite	102,580	987	104	$0.0005
meta-llama/llama-3.1-8b-instruct	166,649	1525	109	$0.0009
meta-llama/llama-4-scout	70,155	479	146	$0.0011
mistralai/mistral-small-2603	3,207	20	160	$0.0349
xiaomi/mimo-v2.5-pro	67,712	274	247	$0.0048
minimax/minimax-m3	120,297	471	255	$0.0023
openai/gpt-5.6-terra	111,389	365	305	$0.0060
openai/gpt-5.6-sol	153,677	412	373	$0.0304
xiaomi/mimo-v2.5	223,313	587	380	$0.0010
microsoft/phi-4	113,671	298	381	$0.0014
nvidia/nemotron-3-ultra-550b-a55b	112,208	229	490	$0.0076
x-ai/grok-4.3	490,642	684	717	$0.0075
openai/gpt-5.6-luna	283,948	390	728	$0.0007
z-ai/glm-5.2	413,174	529	781	$0.0039
openai/gpt-oss-120b	698,041	701	996	$0.0003
openai/gpt-5.5	396,159	363	1091	$0.0379
qwen/qwen3.7-plus	1,407,483	863	1631	$0.0039
qwen/qwen3.7-max	1,282,965	762	1684	$0.0138
google/gemini-3.6-flash	788,423	415	1900	$0.0137
x-ai/grok-4.5	938,537	451	2081	$0.0156
deepseek/deepseek-r1	1,397,958	657	2128	$0.0086
google/gemini-3.5-flash	953,694	439	2172	$0.0162
deepseek/deepseek-r1-0528	1,420,349	650	2185	$0.0072
moonshotai/kimi-k3	874,937	386	2267	$0.0329
deepseek/deepseek-r1-distill-llama-70b	888,615	385	2308	$0.0073
openai/gpt-oss-20b	1,149,006	490	2345	$0.0005
google/gemini-3.1-pro-preview	1,101,910	463	2380	$0.0237
deepseek/deepseek-v4-flash	879,110	362	2428	$0.0012
qwen/qwen3-14b	792,165	318	2491	$0.0044
google/gemini-2.5-pro	1,371,169	512	2678	$0.0210
meituan/longcat-2.0	906,744	265	3422	$0.0061
qwen/qwen3.6-flash	1,468,826	420	3497	$0.0029
qwen/qwen3.5-plus-02-15	2,497,953	700	3569	$0.0050
qwen/qwen3.7-flash	1,477,726	388	3809	$0.0004
deepseek/deepseek-v4-pro	1,099,395	288	3817	$0.0041
qwen/qwen3.6-plus	2,010,766	464	4334	$0.0073
meta/muse-spark-1.1	1,075,701	236	4558	$0.0164
nvidia/nemotron-3-super-120b-a12b	1,508,154	290	5201	$0.0017
openai/o3	741,947	140	5300	$0.0352
deepseek/deepseek-v4-flash-0731	1,339,909	218	6146	$0.0012
inclusionai/ring-2.6-1t	1,286,827	169	7614	$0.0039
qwen/qwen3.5-27b	2,155,828	273	7897	$0.0066
thinkingmachines/inkling-small	2,312,665	260	8895	$0.0122
moonshotai/kimi-k2.6	2,581,920	202	12782	$0.0209
tencent/hy3	2,286,538	171	13372	$0.0038
qwen/qwen3.8-max	1,655,455	114	14522	$0.0890
stepfun/step-3.7-flash	1,508,225	92	16394	$0.0073
openai/o4-mini	1,079,864	52	20767	$0.1059
thinkingmachines/inkling	2,622,481	79	33196	$0.0922
qwen/qwen3-8b	2,139,222	1	2139222	n/a

Cost breakdown (input vs output)

Where each model’s per-episode dollars go, at the same pricing snapshot as every other table. Every model reads the same transcripts, so the input side varies only with the provider’s input price. The output side varies with how much the model writes: a high output share on a modest total usually means reasoning tokens, and a model with a low per-token price can still land mid-table by writing thousands of them. Failed calls are excluded, same as the cost column everywhere else.

Model	Cost / episode	Input	Output	Output share
claude-fable-5	$10.7552	$10.3637	$0.3915	4%
openai/gpt-5.5	$7.6842	$5.3073	$2.3770	31%
openai/gpt-5.6-sol	$6.2293	$5.3073	$0.9221	15%
moonshotai/kimi-k3	$5.9242	$3.2994	$2.6248	44%
claude-opus-5	$5.3761	$5.1818	$0.1942	4%
claude-opus-4-8	$5.3661	$5.1818	$0.1843	3%
claude-opus-4-7	$5.3293	$5.1818	$0.1474	3%
google/gemini-3.1-pro-preview	$4.9160	$2.2714	$2.6446	54%
qwen/qwen3.8-max	$4.2714	$2.2849	$1.9865	47%
google/gemini-2.5-pro	$4.1601	$1.4178	$2.7423	66%
google/gemini-3.5-flash	$3.4202	$1.7036	$1.7166	50%
x-ai/grok-4.5	$3.3554	$2.2292	$1.1262	34%
openai/o3	$3.3100	$2.1229	$1.1871	36%
claude-sonnet-4-6	$3.2386	$3.1091	$0.1295	4%
thinkingmachines/inkling	$3.2258	$1.0626	$2.1631	67%
cohere/command-a	$2.9415	$2.8198	$0.1217	4%
google/gemini-3.6-flash	$2.8863	$1.7037	$1.1826	41%
cohere/command-r-plus-08-2024	$2.8505	$2.8198	$0.0307	1%
qwen/qwen3.7-max	$2.8109	$1.6755	$1.1354	40%
openai/gpt-5.4	$2.7907	$2.6536	$0.1371	5%
meta/muse-spark-1.1	$2.2536	$1.3393	$0.9143	41%
claude-sonnet-5	$2.1452	$2.0727	$0.0724	3%
openai/o4-mini	$2.1179	$1.1676	$0.9503	45%
moonshotai/kimi-k2.6	$1.8854	$0.6254	$1.2600	67%
mistralai/mistral-medium-3-5	$1.8393	$1.7410	$0.0983	5%
x-ai/grok-4.3	$1.6048	$1.3595	$0.2453	15%
deepseek/deepseek-r1	$1.4743	$0.7753	$0.6990	47%
openai/gpt-5.6-terra	$1.1951	$1.0615	$0.1337	11%
deepseek/deepseek-r1-0528	$1.1645	$0.5538	$0.6108	52%
qwen/qwen3.6-plus	$1.1534	$0.3692	$0.7842	68%
claude-haiku-4-5-20251001	$1.0820	$1.0364	$0.0456	4%
qwen/qwen3.5-plus-02-15	$1.0746	$0.2952	$0.7794	73%
thinkingmachines/inkling-small	$1.0333	$0.4782	$0.5550	54%
qwen/qwen3.5-27b	$0.8941	$0.2215	$0.6726	75%
openai/gpt-5.4-mini	$0.8470	$0.7961	$0.0509	6%
nvidia/nemotron-3-ultra-550b-a55b	$0.7782	$0.6974	$0.0808	10%
z-ai/glm-5.2	$0.7376	$0.5961	$0.1414	19%
qwen/qwen3.7-plus	$0.7238	$0.3635	$0.3603	50%
deepseek/deepseek-r1-distill-llama-70b	$0.6913	$0.5492	$0.1422	21%
deepseek/deepseek-v4-pro	$0.6692	$0.4779	$0.1913	29%
mistralai/mistral-large-2512	$0.6184	$0.5803	$0.0381	6%
meituan/longcat-2.0	$0.5573	$0.3396	$0.2176	39%
openai/gpt-3.5-turbo	$0.5564	$0.5395	$0.0168	3%
qwen/qwen3.6-flash	$0.5435	$0.2130	$0.3305	61%
xiaomi/mimo-v2.5-pro	$0.4954	$0.4836	$0.0118	2%
mistralai/mistral-medium-3.1	$0.4847	$0.4643	$0.0204	4%
stepfun/step-3.7-flash	$0.4302	$0.1457	$0.2846	66%
qwen/qwen3-14b	$0.3989	$0.2547	$0.1442	36%
tencent/hy3	$0.3876	$0.1461	$0.2415	62%
google/gemini-2.5-flash	$0.3796	$0.3391	$0.0405	11%
google/gemini-3.5-flash-lite	$0.3643	$0.3407	$0.0235	6%
mistralai/codestral-2508	$0.3585	$0.3482	$0.0103	3%
minimax/minimax-m3	$0.3575	$0.3287	$0.0289	8%
qwen/qwen3-8b	$0.3256	$0.1310	$0.1947	60%
deepseek/deepseek-v3.2	$0.3015	$0.2989	$0.0026	1%
google/gemini-3.1-flash-lite	$0.2988	$0.2837	$0.0151	5%
inclusionai/ring-2.6-1t	$0.2468	$0.0860	$0.1609	65%
meta-llama/llama-4-maverick	$0.2206	$0.2144	$0.0062	3%
nvidia/nemotron-3-super-120b-a12b	$0.2194	$0.0988	$0.1207	55%
deepseek/deepseek-v4-flash	$0.2046	$0.1554	$0.0492	24%
mistralai/mistral-small-2603	$0.1745	$0.1741	$0.0004	0%
xiaomi/mimo-v2.5	$0.1692	$0.1567	$0.0125	7%
deepseek/deepseek-v4-flash-0731	$0.1482	$0.0999	$0.0482	33%
openai/gpt-5.6-luna	$0.1402	$0.1061	$0.0341	24%
google/gemma-4-31b-it	$0.1181	$0.1144	$0.0037	3%
google/gemini-2.5-flash-lite	$0.1132	$0.1050	$0.0082	7%
meta-llama/llama-4-scout	$0.1111	$0.1069	$0.0042	4%
meta-llama/llama-3.3-70b-instruct	$0.1107	$0.1089	$0.0018	2%
qwen/qwen3-235b-a22b-2507	$0.1078	$0.1005	$0.0072	7%
google/gemma-4-26b-a4b-it	$0.0825	$0.0799	$0.0026	3%
microsoft/phi-4	$0.0792	$0.0760	$0.0032	4%
qwen/qwen3.7-flash	$0.0725	$0.0341	$0.0384	53%
openai/gpt-oss-120b	$0.0634	$0.0397	$0.0237	37%
openai/gpt-oss-20b	$0.0620	$0.0322	$0.0299	48%
meta-llama/llama-3.1-8b-instruct	$0.0572	$0.0545	$0.0027	5%

Trial variance (determinism check)

All trials run at temperature 0.0. If a model produces stable output you’d expect the F1 stdev across trials to be near zero. Higher numbers mean the model is non-deterministic even at temp=0. That’s fine to know, but means you cannot trust a single trial’s number for that model.

Model	Mean F1 stdev across episodes	Highest single-episode stdev
claude-haiku-4-5-20251001	0.0415	0.1845
claude-sonnet-4-6	0.0386	0.2236
x-ai/grok-4.5	0.0190	0.0894
google/gemini-3.6-flash	0.0114	0.0596
x-ai/grok-4.3	0.0631	0.1675
google/gemini-3.5-flash	0.0090	0.0555
mistralai/mistral-medium-3-5	0.0239	0.1082
openai/gpt-5.5	0.0259	0.0994
claude-fable-5	0.0456	0.2449
qwen/qwen3.5-plus-02-15	0.0302	0.1118
google/gemini-3.5-flash-lite	0.0738	0.2226
openai/gpt-5.6-terra	0.0559	0.1482
claude-opus-4-8	0.0372	0.1565
openai/gpt-5.6-luna	0.0440	0.1571
google/gemini-3.1-pro-preview	0.0254	0.1118
claude-opus-5	0.0113	0.0943
google/gemini-3.1-flash-lite	0.0278	0.1118
qwen/qwen3.6-flash	0.0842	0.2300
qwen/qwen3.7-max	0.0505	0.1873
openai/gpt-5.6-sol	0.0733	0.2739
claude-sonnet-5	0.0904	0.2191
claude-opus-4-7	0.0604	0.2145
deepseek/deepseek-v4-flash	0.1230	0.4472
qwen/qwen3.7-flash	0.1210	0.3742
moonshotai/kimi-k3	0.0886	0.2981
google/gemini-2.5-pro	0.0513	0.1659
mistralai/mistral-medium-3.1	0.0462	0.1325
google/gemini-2.5-flash	0.0131	0.1164
openai/gpt-oss-120b	0.1260	0.2454
deepseek/deepseek-v4-pro	0.1624	0.2733
google/gemma-4-31b-it	0.1375	0.5055
openai/gpt-5.4	0.0909	0.3118
minimax/minimax-m3	0.2009	0.4472
google/gemini-2.5-flash-lite	0.0704	0.1789
qwen/qwen3.6-plus	0.1480	0.3635
z-ai/glm-5.2	0.1169	0.1789
deepseek/deepseek-v4-flash-0731	0.1872	0.5477
deepseek/deepseek-v3.2	0.0721	0.3173
meta/muse-spark-1.1	0.2236	0.4346
openai/gpt-5.4-mini	0.0753	0.2236
nvidia/nemotron-3-super-120b-a12b	0.1111	0.2549
deepseek/deepseek-r1	0.1403	0.3651
google/gemma-4-26b-a4b-it	0.1361	0.2981
qwen/qwen3.5-27b	0.2144	0.4336
mistralai/mistral-large-2512	0.0434	0.1491
deepseek/deepseek-r1-0528	0.1604	0.3578
openai/gpt-oss-20b	0.1630	0.4472
meta-llama/llama-4-maverick	0.0330	0.1789
qwen/qwen3.7-plus	0.0626	0.2454
mistralai/codestral-2508	0.0691	0.2236
qwen/qwen3-235b-a22b-2507	0.1504	0.3249
meta-llama/llama-3.3-70b-instruct	0.1166	0.3651
tencent/hy3	0.1238	0.2460
xiaomi/mimo-v2.5	0.1504	0.3249
openai/o3	0.1878	0.4714
nvidia/nemotron-3-ultra-550b-a55b	0.2311	0.4714
meta-llama/llama-4-scout	0.2238	0.5477
xiaomi/mimo-v2.5-pro	0.2166	0.3651
cohere/command-a	0.0519	0.2140
moonshotai/kimi-k2.6	0.1521	0.2888
stepfun/step-3.7-flash	0.1555	0.4346
meituan/longcat-2.0	0.1640	0.4714
deepseek/deepseek-r1-distill-llama-70b	0.1088	0.3789
qwen/qwen3-14b	0.1158	0.1966
cohere/command-r-plus-08-2024	0.1484	0.3651
thinkingmachines/inkling-small	0.1666	0.3138
inclusionai/ring-2.6-1t	0.1395	0.2828
openai/gpt-3.5-turbo	0.0208	0.0994
microsoft/phi-4	0.1084	0.3651
qwen/qwen3.8-max	0.1118	0.2844
meta-llama/llama-3.1-8b-instruct	0.1645	0.3789
thinkingmachines/inkling	0.1532	0.4561
openai/o4-mini	0.1817	0.3651
mistralai/mistral-small-2603	0.0000	0.0000
qwen/qwen3-8b	0.0000	0.0000

Cross-model agreement

For each of the 171 (episode, window, trial-equivalent) entries, how many of the 75 active models predicted at least one ad? High-agreement windows are unambiguous ads (or unambiguously not ads). Low-agreement windows are where individual models disagree, and are candidates for ensemble voting if you want a cheap accuracy boost.

Models predicting an ad	Window count	Share
5 of 75	1	0.6%
7 of 75	8	4.7%
8 of 75	4	2.3%
9 of 75	6	3.5%
10 of 75	12	7.0%
11 of 75	7	4.1%
12 of 75	4	2.3%
13 of 75	9	5.3%
14 of 75	4	2.3%
15 of 75	6	3.5%
16 of 75	10	5.8%
17 of 75	4	2.3%
18 of 75	6	3.5%
20 of 75	2	1.2%
22 of 75	4	2.3%
23 of 75	1	0.6%
24 of 75	1	0.6%
25 of 75	1	0.6%
26 of 75	1	0.6%
27 of 75	1	0.6%
29 of 75	1	0.6%
32 of 75	1	0.6%
35 of 75	1	0.6%
39 of 75	1	0.6%
40 of 75	1	0.6%
44 of 75	1	0.6%
46 of 75	2	1.2%
51 of 75	1	0.6%
52 of 75	1	0.6%
53 of 75	1	0.6%
59 of 75	1	0.6%
61 of 75	2	1.2%
62 of 75	2	1.2%
63 of 75	4	2.3%
64 of 75	5	2.9%
65 of 75	5	2.9%
66 of 75	11	6.4%
67 of 75	9	5.3%
68 of 75	10	5.8%
69 of 75	7	4.1%
70 of 75	7	4.1%
71 of 75	4	2.3%
72 of 75	1	0.6%

Read this as: rows near the top are windows where the field disagrees (most models said no, a few said yes, usually false positives); rows near the bottom are windows where the field broadly agrees (typical of clear sponsor reads).

Per-model alignment with consensus

Same data, viewed per model. For each window, the majority is whether more than half of the 75 active models flagged an ad. Then for each model: did it vote with the majority or against it? Four buckets:

• with-yes: this model voted yes, majority also voted yes (likely true positive)
• with-no: this model voted no, majority also voted no (likely true negative)
• broke-yes: this model voted yes, majority voted no (likely false positive / hallucination)
• broke-no: this model voted no, majority voted yes (likely missed real ad)

Alignment rate is (with-yes + with-no) / total. High alignment means the model tracks the consensus; low alignment means it disagrees often, which could be brilliance or noise depending on whether its disagreements are also where its F1 wins or loses.

Model	with-yes	with-no	broke-yes	broke-no	Alignment
claude-opus-5	75	93	2	1	98.2%
claude-opus-4-8	74	93	2	2	97.7%
claude-fable-5	74	92	3	2	97.1%
claude-sonnet-5	72	94	1	4	97.1%
google/gemini-2.5-flash	75	91	4	1	97.1%
google/gemini-3.1-pro-preview	72	94	1	4	97.1%
openai/gpt-5.5	75	91	4	1	97.1%
qwen/qwen3.5-plus-02-15	71	95	0	5	97.1%
qwen/qwen3.6-flash	73	93	2	3	97.1%
qwen/qwen3.7-max	72	94	1	4	97.1%
x-ai/grok-4.3	72	94	1	4	97.1%
claude-opus-4-7	74	91	4	2	96.5%
google/gemini-3.5-flash	71	94	1	5	96.5%
google/gemini-3.6-flash	71	94	1	5	96.5%
openai/gpt-oss-120b	76	89	6	0	96.5%
qwen/qwen3.5-27b	72	93	2	4	96.5%
qwen/qwen3.7-flash	72	93	2	4	96.5%
deepseek/deepseek-v4-flash	74	90	5	2	95.9%
moonshotai/kimi-k3	73	91	4	3	95.9%
x-ai/grok-4.5	70	94	1	6	95.9%
claude-haiku-4-5-20251001	70	93	2	6	95.3%
claude-sonnet-4-6	69	94	1	7	95.3%
google/gemma-4-31b-it	75	88	7	1	95.3%
meta/muse-spark-1.1	68	95	0	8	95.3%
openai/gpt-5.6-luna	76	87	8	0	95.3%
google/gemini-2.5-pro	76	86	9	0	94.7%
deepseek/deepseek-v4-flash-0731	65	95	0	11	93.6%
google/gemma-4-26b-a4b-it	75	84	11	1	93.0%
qwen/qwen3.6-plus	70	89	6	6	93.0%
nvidia/nemotron-3-super-120b-a12b	66	91	4	10	91.8%
openai/gpt-5.6-terra	74	83	12	2	91.8%
openai/gpt-oss-20b	73	84	11	3	91.8%
openai/o3	62	95	0	14	91.8%
google/gemini-3.5-flash-lite	62	94	1	14	91.2%
meta-llama/llama-3.3-70b-instruct	70	86	9	6	91.2%
mistralai/mistral-medium-3-5	67	89	6	9	91.2%
google/gemini-3.1-flash-lite	76	79	16	0	90.6%
mistralai/mistral-medium-3.1	61	93	2	15	90.1%
qwen/qwen3-14b	72	82	13	4	90.1%
deepseek/deepseek-v4-pro	71	81	14	5	88.9%
inclusionai/ring-2.6-1t	61	91	4	15	88.9%
minimax/minimax-m3	65	85	10	11	87.7%
meta-llama/llama-4-maverick	76	72	23	0	86.5%
tencent/hy3	55	93	2	21	86.5%
xiaomi/mimo-v2.5-pro	71	74	21	5	84.8%
deepseek/deepseek-r1-distill-llama-70b	54	90	5	22	84.2%
meta-llama/llama-4-scout	71	73	22	5	84.2%
meituan/longcat-2.0	48	95	0	28	83.6%
deepseek/deepseek-v3.2	44	95	0	32	81.3%
google/gemini-2.5-flash-lite	75	59	36	1	78.4%
deepseek/deepseek-r1	76	56	39	0	77.2%
microsoft/phi-4	41	91	4	35	77.2%
openai/o4-mini	38	94	1	38	77.2%
openai/gpt-5.6-sol	75	55	40	1	76.0%
thinkingmachines/inkling	41	88	7	35	75.4%
meta-llama/llama-3.1-8b-instruct	70	57	38	6	74.3%
mistralai/codestral-2508	67	60	35	9	74.3%
stepfun/step-3.7-flash	32	95	0	44	74.3%
cohere/command-r-plus-08-2024	41	84	11	35	73.1%
openai/gpt-5.4	76	47	48	0	71.9%
qwen/qwen3.8-max	25	95	0	51	70.2%
nvidia/nemotron-3-ultra-550b-a55b	68	51	44	8	69.6%
deepseek/deepseek-r1-0528	75	42	53	1	68.4%
mistralai/mistral-large-2512	75	34	61	1	63.7%
cohere/command-a	75	27	68	1	59.6%
moonshotai/kimi-k2.6	47	52	43	29	57.9%
thinkingmachines/inkling-small	60	39	56	16	57.9%
openai/gpt-5.4-mini	76	22	73	0	57.3%
z-ai/glm-5.2	75	23	72	1	57.3%
mistralai/mistral-small-2603	1	95	0	75	56.1%
qwen/qwen3-8b	1	95	0	75	56.1%
openai/gpt-3.5-turbo	75	13	82	1	51.5%
qwen/qwen3-235b-a22b-2507	74	14	81	2	51.5%
qwen/qwen3.7-plus	75	5	90	1	46.8%
xiaomi/mimo-v2.5	76	3	92	0	46.2%

Windows flagged with no truth ad

The other side of the histogram: windows the ground truth marks ad-free, ranked by how many of the 75 models flagged them anyway (in at least one trial). A window near the top is either content that genuinely resembles an ad, which is what precision-focused validator rules should train against, or a spot the truth file missed. Either way these are the first windows worth a manual re-listen; on a corpus this size a single mislabeled window moves scores. No-ad control episodes are included and tagged.

Episode	Window	Span	Models flagging
ep-it-s-a-thing-e339179dfad6	0	0-600s	67 of 75
ep-on-air-with-dan-and-alex2-574e4f303730	6	2520-3120s	59 of 75
ep-on-air-with-dan-and-alex2-574e4f303730	5	2100-2700s	53 of 75
ep-drink-champs-30c9a2d49f13	35	14700-15300s	52 of 75
ep-the-brilliant-idiots-0bb9bf634c8e	10	4200-4800s	51 of 75
ep-drink-champs-30c9a2d49f13	12	5040-5640s	46 of 75
ep-the-brilliant-idiots-0bb9bf634c8e	9	3780-4380s	46 of 75
ep-daily-gist-chicago-70a82fe93a5c	2	840-1271s	40 of 75
ep-daily-gist-chicago-70a82fe93a5c	3	1260-1271s	39 of 75
ep-the-brilliant-idiots-0bb9bf634c8e	14	5880-6480s	32 of 75
ep-ai-cloud-essentials-e8dc897fbd6b (no-ad control)	0	0-600s	29 of 75
ep-crime-junkie-8ce498f299d7	1	420-1020s	27 of 75
ep-security-now-audio-2850b24903b2	11	4620-5220s	26 of 75
ep-on-air-with-dan-and-alex2-574e4f303730	2	840-1440s	25 of 75
ep-the-brilliant-idiots-0bb9bf634c8e	3	1260-1860s	24 of 75

… and 88 more with 2+ votes.

Detection rate by ad characteristic

Aggregate detection rates often hide systematic blind spots. Below: for each model, what fraction of truth ads in each bucket were detected (matched at IoU >= 0.5).

By ad length

Truth ads bucketed by duration: short (<30s), medium (30-90s), long (>=90s). Cell values are detection rate (fraction of truth ads in that bucket the model caught), with the sample size n so a misleading 1.00 on a 2-ad bucket doesn’t get over-weighted. Models that systematically miss short ads usually fail on network-inserted brand-tagline spots; missing long ads is rarer and usually means the model gave up before processing the full window.

Model	long (>=90s)	medium (30-90s)	short (<30s)
claude-fable-5	0.90 (n=140)	0.85 (n=75)	0.70 (n=40)
claude-haiku-4-5-20251001	0.88 (n=140)	0.88 (n=75)	0.85 (n=40)
claude-opus-4-7	0.97 (n=140)	0.72 (n=75)	0.45 (n=40)
claude-opus-4-8	0.92 (n=140)	0.79 (n=75)	0.57 (n=40)
claude-opus-5	0.96 (n=140)	0.81 (n=75)	0.50 (n=40)
claude-sonnet-4-6	0.89 (n=140)	0.87 (n=75)	0.53 (n=40)
claude-sonnet-5	0.91 (n=140)	0.76 (n=75)	0.47 (n=40)
cohere/command-a	0.47 (n=140)	0.57 (n=75)	0.62 (n=40)
cohere/command-r-plus-08-2024	0.48 (n=140)	0.19 (n=75)	0.17 (n=40)
deepseek/deepseek-r1	0.66 (n=140)	0.68 (n=75)	0.70 (n=40)
deepseek/deepseek-r1-0528	0.60 (n=140)	0.67 (n=75)	0.70 (n=40)
deepseek/deepseek-r1-distill-llama-70b	0.35 (n=140)	0.25 (n=75)	0.68 (n=40)
deepseek/deepseek-v3.2	0.71 (n=140)	0.44 (n=75)	0.17 (n=40)
deepseek/deepseek-v4-flash	0.75 (n=140)	0.59 (n=75)	0.65 (n=40)
deepseek/deepseek-v4-flash-0731	0.50 (n=140)	0.51 (n=75)	0.50 (n=40)
deepseek/deepseek-v4-pro	0.69 (n=140)	0.59 (n=75)	0.53 (n=40)
google/gemini-2.5-flash	0.89 (n=140)	0.87 (n=75)	0.78 (n=40)
google/gemini-2.5-flash-lite	0.83 (n=140)	0.76 (n=75)	0.85 (n=40)
google/gemini-2.5-pro	0.86 (n=140)	0.71 (n=75)	0.60 (n=40)
google/gemini-3.1-flash-lite	0.89 (n=140)	0.93 (n=75)	0.88 (n=40)
google/gemini-3.1-pro-preview	0.89 (n=140)	0.83 (n=75)	0.50 (n=40)
google/gemini-3.5-flash	0.89 (n=140)	0.81 (n=75)	0.62 (n=40)
google/gemini-3.5-flash-lite	0.87 (n=140)	0.80 (n=75)	0.57 (n=40)
google/gemini-3.6-flash	0.89 (n=140)	0.85 (n=75)	0.55 (n=40)
google/gemma-4-26b-a4b-it	0.50 (n=140)	0.67 (n=75)	0.88 (n=40)
google/gemma-4-31b-it	0.79 (n=140)	0.65 (n=75)	0.47 (n=40)
inclusionai/ring-2.6-1t	0.17 (n=140)	0.32 (n=75)	0.40 (n=40)
meituan/longcat-2.0	0.36 (n=140)	0.35 (n=75)	0.38 (n=40)
meta-llama/llama-3.1-8b-instruct	0.23 (n=140)	0.33 (n=75)	0.20 (n=40)
meta-llama/llama-3.3-70b-instruct	0.40 (n=140)	0.55 (n=75)	0.57 (n=40)
meta-llama/llama-4-maverick	0.59 (n=140)	0.71 (n=75)	0.53 (n=40)
meta-llama/llama-4-scout	0.43 (n=140)	0.40 (n=75)	0.38 (n=40)
meta/muse-spark-1.1	0.64 (n=140)	0.40 (n=75)	0.42 (n=40)
microsoft/phi-4	0.14 (n=140)	0.35 (n=75)	0.25 (n=40)
minimax/minimax-m3	0.64 (n=140)	0.60 (n=75)	0.53 (n=40)
mistralai/codestral-2508	0.71 (n=140)	0.67 (n=75)	0.35 (n=40)
mistralai/mistral-large-2512	0.89 (n=140)	0.85 (n=75)	0.70 (n=40)
mistralai/mistral-medium-3-5	0.89 (n=140)	0.81 (n=75)	0.88 (n=40)
mistralai/mistral-medium-3.1	0.78 (n=140)	0.75 (n=75)	0.75 (n=40)
mistralai/mistral-small-2603	0.00 (n=140)	0.00 (n=75)	0.12 (n=40)
moonshotai/kimi-k2.6	0.33 (n=140)	0.47 (n=75)	0.23 (n=40)
moonshotai/kimi-k3	0.71 (n=140)	0.77 (n=75)	0.55 (n=40)
nvidia/nemotron-3-super-120b-a12b	0.44 (n=140)	0.56 (n=75)	0.70 (n=40)
nvidia/nemotron-3-ultra-550b-a55b	0.39 (n=140)	0.40 (n=75)	0.45 (n=40)
openai/gpt-3.5-turbo	0.34 (n=140)	0.45 (n=75)	0.75 (n=40)
openai/gpt-5.4	0.90 (n=140)	0.68 (n=75)	0.55 (n=40)
openai/gpt-5.4-mini	0.85 (n=140)	0.59 (n=75)	0.70 (n=40)
openai/gpt-5.5	0.89 (n=140)	0.76 (n=75)	0.55 (n=40)
openai/gpt-5.6-luna	0.86 (n=140)	0.72 (n=75)	0.70 (n=40)
openai/gpt-5.6-sol	0.86 (n=140)	0.76 (n=75)	0.70 (n=40)
openai/gpt-5.6-terra	0.84 (n=140)	0.75 (n=75)	0.62 (n=40)
openai/gpt-oss-120b	0.69 (n=140)	0.81 (n=75)	0.85 (n=40)
openai/gpt-oss-20b	0.49 (n=140)	0.57 (n=75)	0.65 (n=40)
openai/o3	0.46 (n=140)	0.28 (n=75)	0.23 (n=40)
openai/o4-mini	0.06 (n=140)	0.09 (n=75)	0.12 (n=40)
qwen/qwen3-14b	0.20 (n=140)	0.47 (n=75)	0.68 (n=40)
qwen/qwen3-235b-a22b-2507	0.54 (n=140)	0.64 (n=75)	0.65 (n=40)
qwen/qwen3-8b	0.00 (n=140)	0.00 (n=75)	0.00 (n=40)
qwen/qwen3.5-27b	0.61 (n=140)	0.41 (n=75)	0.47 (n=40)
qwen/qwen3.5-plus-02-15	0.91 (n=140)	0.81 (n=75)	0.62 (n=40)
qwen/qwen3.6-flash	0.80 (n=140)	0.71 (n=75)	0.62 (n=40)
qwen/qwen3.6-plus	0.61 (n=140)	0.64 (n=75)	0.65 (n=40)
qwen/qwen3.7-flash	0.74 (n=140)	0.71 (n=75)	0.60 (n=40)
qwen/qwen3.7-max	0.79 (n=140)	0.91 (n=75)	0.62 (n=40)
qwen/qwen3.7-plus	0.76 (n=140)	0.73 (n=75)	0.62 (n=40)
qwen/qwen3.8-max	0.17 (n=140)	0.29 (n=75)	0.05 (n=40)
stepfun/step-3.7-flash	0.23 (n=95)	0.35 (n=65)	0.40 (n=35)
tencent/hy3	0.39 (n=140)	0.47 (n=75)	0.35 (n=40)
thinkingmachines/inkling	0.13 (n=119)	0.22 (n=73)	0.11 (n=38)
thinkingmachines/inkling-small	0.29 (n=140)	0.43 (n=75)	0.33 (n=40)
x-ai/grok-4.3	0.86 (n=140)	0.85 (n=75)	0.72 (n=40)
x-ai/grok-4.5	0.89 (n=140)	0.87 (n=75)	0.62 (n=40)
xiaomi/mimo-v2.5	0.70 (n=140)	0.57 (n=75)	0.53 (n=40)
xiaomi/mimo-v2.5-pro	0.48 (n=140)	0.31 (n=75)	0.33 (n=40)
z-ai/glm-5.2	0.80 (n=140)	0.73 (n=75)	0.55 (n=40)

By ad position

Truth ads bucketed by where they fall in the episode: pre-roll (first 10%), mid-roll (10-90%), post-roll (last 10%). Cell values are the same detection-rate-with-n format as ad length. A common failure pattern in our data: most models detect pre-roll and mid-roll reliably and miss post-roll, because the prompt windows near the end often catch the model mid-reasoning or with fewer transition phrases to anchor on.

Model	pre-roll (<10%)	mid-roll (10-90%)	post-roll (>90%)
claude-fable-5	0.80 (n=75)	0.94 (n=125)	0.73 (n=55)
claude-haiku-4-5-20251001	0.80 (n=75)	0.97 (n=125)	0.76 (n=55)
claude-opus-4-7	0.76 (n=75)	0.86 (n=125)	0.80 (n=55)
claude-opus-4-8	0.80 (n=75)	0.91 (n=125)	0.67 (n=55)
claude-opus-5	0.85 (n=75)	0.84 (n=125)	0.84 (n=55)
claude-sonnet-4-6	0.73 (n=75)	0.93 (n=125)	0.73 (n=55)
claude-sonnet-5	0.79 (n=75)	0.87 (n=125)	0.64 (n=55)
cohere/command-a	0.49 (n=75)	0.57 (n=125)	0.47 (n=55)
cohere/command-r-plus-08-2024	0.25 (n=75)	0.48 (n=125)	0.16 (n=55)
deepseek/deepseek-r1	0.65 (n=75)	0.74 (n=125)	0.55 (n=55)
deepseek/deepseek-r1-0528	0.63 (n=75)	0.68 (n=125)	0.55 (n=55)
deepseek/deepseek-r1-distill-llama-70b	0.29 (n=75)	0.39 (n=125)	0.44 (n=55)
deepseek/deepseek-v3.2	0.63 (n=75)	0.66 (n=125)	0.18 (n=55)
deepseek/deepseek-v4-flash	0.61 (n=75)	0.78 (n=125)	0.58 (n=55)
deepseek/deepseek-v4-flash-0731	0.48 (n=75)	0.53 (n=125)	0.47 (n=55)
deepseek/deepseek-v4-pro	0.55 (n=75)	0.73 (n=125)	0.55 (n=55)
google/gemini-2.5-flash	0.73 (n=75)	0.96 (n=125)	0.82 (n=55)
google/gemini-2.5-flash-lite	0.68 (n=75)	0.90 (n=125)	0.78 (n=55)
google/gemini-2.5-pro	0.68 (n=75)	0.87 (n=125)	0.69 (n=55)
google/gemini-3.1-flash-lite	0.80 (n=75)	0.99 (n=125)	0.82 (n=55)
google/gemini-3.1-pro-preview	0.80 (n=75)	0.86 (n=125)	0.71 (n=55)
google/gemini-3.5-flash	0.75 (n=75)	0.92 (n=125)	0.73 (n=55)
google/gemini-3.5-flash-lite	0.79 (n=75)	0.88 (n=125)	0.65 (n=55)
google/gemini-3.6-flash	0.80 (n=75)	0.89 (n=125)	0.73 (n=55)
google/gemma-4-26b-a4b-it	0.57 (n=75)	0.69 (n=125)	0.47 (n=55)
google/gemma-4-31b-it	0.68 (n=75)	0.75 (n=125)	0.60 (n=55)
inclusionai/ring-2.6-1t	0.21 (n=75)	0.30 (n=125)	0.20 (n=55)
meituan/longcat-2.0	0.28 (n=75)	0.44 (n=125)	0.27 (n=55)
meta-llama/llama-3.1-8b-instruct	0.17 (n=75)	0.28 (n=125)	0.31 (n=55)
meta-llama/llama-3.3-70b-instruct	0.37 (n=75)	0.52 (n=125)	0.49 (n=55)
meta-llama/llama-4-maverick	0.59 (n=75)	0.70 (n=125)	0.45 (n=55)
meta-llama/llama-4-scout	0.32 (n=75)	0.49 (n=125)	0.36 (n=55)
meta/muse-spark-1.1	0.47 (n=75)	0.64 (n=125)	0.40 (n=55)
microsoft/phi-4	0.31 (n=75)	0.10 (n=125)	0.36 (n=55)
minimax/minimax-m3	0.43 (n=75)	0.71 (n=125)	0.62 (n=55)
mistralai/codestral-2508	0.37 (n=75)	0.75 (n=125)	0.75 (n=55)
mistralai/mistral-large-2512	0.79 (n=75)	0.94 (n=125)	0.73 (n=55)
mistralai/mistral-medium-3-5	0.73 (n=75)	0.97 (n=125)	0.82 (n=55)
mistralai/mistral-medium-3.1	0.80 (n=75)	0.78 (n=125)	0.67 (n=55)
mistralai/mistral-small-2603	0.00 (n=75)	0.04 (n=125)	0.00 (n=55)
moonshotai/kimi-k2.6	0.36 (n=75)	0.31 (n=125)	0.44 (n=55)
moonshotai/kimi-k3	0.71 (n=75)	0.73 (n=125)	0.65 (n=55)
nvidia/nemotron-3-super-120b-a12b	0.52 (n=75)	0.55 (n=125)	0.44 (n=55)
nvidia/nemotron-3-ultra-550b-a55b	0.31 (n=75)	0.44 (n=125)	0.44 (n=55)
openai/gpt-3.5-turbo	0.39 (n=75)	0.48 (n=125)	0.42 (n=55)
openai/gpt-5.4	0.72 (n=75)	0.86 (n=125)	0.67 (n=55)
openai/gpt-5.4-mini	0.67 (n=75)	0.90 (n=125)	0.53 (n=55)
openai/gpt-5.5	0.80 (n=75)	0.86 (n=125)	0.65 (n=55)
openai/gpt-5.6-luna	0.80 (n=75)	0.90 (n=125)	0.56 (n=55)
openai/gpt-5.6-sol	0.79 (n=75)	0.88 (n=125)	0.65 (n=55)
openai/gpt-5.6-terra	0.73 (n=75)	0.87 (n=125)	0.64 (n=55)
openai/gpt-oss-120b	0.69 (n=75)	0.79 (n=125)	0.73 (n=55)
openai/gpt-oss-20b	0.45 (n=75)	0.61 (n=125)	0.49 (n=55)
openai/o3	0.24 (n=75)	0.45 (n=125)	0.36 (n=55)
openai/o4-mini	0.08 (n=75)	0.08 (n=125)	0.07 (n=55)
qwen/qwen3-14b	0.33 (n=75)	0.37 (n=125)	0.35 (n=55)
qwen/qwen3-235b-a22b-2507	0.44 (n=75)	0.66 (n=125)	0.62 (n=55)
qwen/qwen3-8b	0.00 (n=75)	0.00 (n=125)	0.00 (n=55)
qwen/qwen3.5-27b	0.45 (n=75)	0.59 (n=125)	0.49 (n=55)
qwen/qwen3.5-plus-02-15	0.80 (n=75)	0.90 (n=125)	0.75 (n=55)
qwen/qwen3.6-flash	0.67 (n=75)	0.84 (n=125)	0.64 (n=55)
qwen/qwen3.6-plus	0.60 (n=75)	0.67 (n=125)	0.55 (n=55)
qwen/qwen3.7-flash	0.67 (n=75)	0.77 (n=125)	0.64 (n=55)
qwen/qwen3.7-max	0.72 (n=75)	0.86 (n=125)	0.75 (n=55)
qwen/qwen3.7-plus	0.77 (n=75)	0.76 (n=125)	0.60 (n=55)
qwen/qwen3.8-max	0.27 (n=75)	0.08 (n=125)	0.33 (n=55)
stepfun/step-3.7-flash	0.36 (n=55)	0.32 (n=95)	0.20 (n=45)
tencent/hy3	0.32 (n=75)	0.47 (n=125)	0.36 (n=55)
thinkingmachines/inkling	0.10 (n=69)	0.15 (n=109)	0.23 (n=52)
thinkingmachines/inkling-small	0.31 (n=75)	0.32 (n=125)	0.40 (n=55)
x-ai/grok-4.3	0.73 (n=75)	0.92 (n=125)	0.78 (n=55)
x-ai/grok-4.5	0.80 (n=75)	0.92 (n=125)	0.73 (n=55)
xiaomi/mimo-v2.5	0.49 (n=75)	0.72 (n=125)	0.64 (n=55)
xiaomi/mimo-v2.5-pro	0.35 (n=75)	0.46 (n=125)	0.35 (n=55)
z-ai/glm-5.2	0.68 (n=75)	0.85 (n=125)	0.58 (n=55)

Quick Comparison

One row per model, one column per episode. The headline columns (F1, Cost/ep, p50) summarize across all episodes; the per-episode columns let you see whether a model’s average hides wide swings (a model that scores well overall might still bomb on a specific genre). The right-most F1 stdev column averages the per-trial standard deviations across episodes; high values mean the model isn’t deterministic at temperature 0.0, so its single-trial F1 number is noisy. Moderation blocked is the share of attempted calls the provider refused on content grounds; those windows never reach scoring, so any non-zero value means that row’s F1 was computed on a subset of the corpus and is not comparable to a row at -.

Model	F1	Cost/ep	p50	ep-crime-junkie-8ce498f299d7	ep-daily-gist-chicago-70a82fe93a5c	ep-daily-tech-news-show-b576979e1fe8	ep-daily-tech-news-show-c1904b8605f7	ep-drink-champs-30c9a2d49f13	ep-glt1412515089-373d5ba5007b	ep-it-s-a-thing-e339179dfad6	ep-on-air-with-dan-and-alex2-574e4f303730	ep-security-now-audio-2850b24903b2	ep-the-brilliant-idiots-0bb9bf634c8e	ep-the-tim-dillon-show-f62bd5fa1cfe	ep-tosh-show-5f6894439bb6	ep-ai-cloud-essentials-e8dc897fbd6b (no-ad)	ep-oxide-and-friends-ce789ff5b62e (no-ad)	F1 stdev	Moderation blocked
claude-haiku-4-5-20251001	0.842	$1.0820	24.2s	1.000	1.000	1.000	0.567	0.937	0.814	0.667	0.920	0.714	1.000	0.812	0.669	PASS	PASS	0.041	-
claude-sonnet-4-6	0.799	$3.2386	4.1s	1.000	0.600	1.000	0.556	0.891	0.857	0.667	0.773	0.695	1.000	0.667	0.879	PASS	PASS	0.039	-
x-ai/grok-4.5	0.797	$3.3554	10.9s	0.956	0.500	1.000	0.509	0.899	0.857	0.667	0.960	0.714	0.857	0.800	0.848	PASS	PASS	0.019	-
google/gemini-3.6-flash	0.796	$2.8863	4.5s	1.000	0.500	1.000	0.589	0.842	0.857	0.667	0.773	0.667	1.000	0.800	0.857	PASS	PASS	0.011	-
x-ai/grok-4.3	0.789	$1.6048	3.7s	1.000	0.460	0.978	0.578	0.814	0.921	0.667	1.000	0.714	0.743	0.842	0.745	PASS	PASS	0.063	-
google/gemini-3.5-flash	0.787	$3.4202	5.6s	1.000	0.500	1.000	0.578	0.873	0.857	0.667	0.800	0.714	1.000	0.800	0.656	PASS	PASS	0.009	-
mistralai/mistral-medium-3-5	0.787	$1.8393	1.1s	1.000	0.800	1.000	0.589	0.833	0.740	0.667	1.000	0.714	0.667	0.716	0.715	PASS	PASS	0.024	-
openai/gpt-5.5	0.784	$7.6842	5.4s	1.000	0.400	1.000	0.667	0.823	0.914	0.667	0.800	0.769	0.943	0.800	0.622	FAIL (1 FP)	PASS	0.026	-
claude-fable-5	0.783	$10.7552	6.6s	0.933	0.700	0.978	0.529	0.770	0.857	0.667	0.960	0.667	0.667	0.800	0.864	PASS	PASS	0.046	-
qwen/qwen3.5-plus-02-15	0.779	$1.0746	29.2s	1.000	0.500	0.889	0.545	0.967	0.857	0.667	0.960	0.686	0.800	0.748	0.727	PASS	PASS	0.030	-
google/gemini-3.5-flash-lite	0.776	$0.3643	0.7s	0.876	1.000	0.892	0.502	0.959	0.531	0.667	0.933	0.695	0.700	0.825	0.727	PASS	PASS	0.074	-
openai/gpt-5.6-terra	0.768	$1.1951	2.1s	1.000	0.400	0.950	0.653	0.774	0.857	0.600	0.960	0.716	0.791	0.800	0.711	FAIL (2 FP)	FAIL (1 FP)	0.056	-
claude-opus-4-8	0.760	$5.3661	7.8s	0.889	0.800	0.956	0.527	0.870	0.743	0.667	0.800	0.695	0.556	0.800	0.822	PASS	PASS	0.037	-
openai/gpt-5.6-luna	0.754	$0.1402	3.5s	0.933	0.400	1.000	0.609	0.778	0.949	0.667	0.693	0.758	0.793	0.800	0.667	PASS	PASS	0.044	-
google/gemini-3.1-pro-preview	0.752	$4.9160	8.3s	1.000	0.500	0.950	0.578	0.834	0.857	0.633	0.667	0.667	0.857	0.800	0.676	PASS	PASS	0.025	-
claude-opus-5	0.748	$5.3761	3.7s	1.000	0.400	1.000	0.564	0.863	0.857	0.500	0.800	0.857	0.667	0.800	0.667	PASS	PASS	0.011	-
google/gemini-3.1-flash-lite	0.747	$0.2988	0.8s	0.950	0.800	0.956	0.567	0.858	0.681	0.667	0.800	0.588	0.540	0.833	0.727	FAIL (1 FP)	PASS	0.028	-
qwen/qwen3.6-flash	0.747	$0.5435	7.6s	1.000	0.500	0.921	0.602	0.641	0.755	0.733	0.840	0.714	0.810	0.720	0.723	PASS	PASS	0.084	-
qwen/qwen3.7-max	0.746	$2.8109	23.4s	0.950	0.500	1.000	0.462	0.628	0.857	0.667	0.840	0.695	0.821	0.848	0.681	PASS	PASS	0.051	-
openai/gpt-5.6-sol	0.738	$6.2293	3.7s	1.000	0.420	1.000	0.600	0.648	0.739	0.467	0.747	0.736	0.834	0.773	0.889	FAIL (3 FP)	FAIL (2 FP)	0.073	-
claude-sonnet-5	0.732	$2.1452	8.5s	0.938	0.560	0.911	0.492	0.806	0.857	0.633	0.960	0.693	0.503	0.785	0.647	PASS	PASS	0.090	-
claude-opus-4-7	0.731	$5.3293	4.6s	0.978	0.400	1.000	0.767	0.825	0.857	0.667	0.740	0.693	0.578	0.727	0.538	PASS	PASS	0.060	-
deepseek/deepseek-v4-flash	0.708	$0.2046	6.3s	1.000	0.673	0.886	0.402	0.687	0.921	0.467	0.800	0.673	0.903	0.733	0.355	PASS	PASS	0.123	-
qwen/qwen3.7-flash	0.706	$0.0725	10.1s	1.000	0.500	0.771	0.602	0.546	0.664	0.600	0.800	0.714	0.769	0.785	0.718	PASS	PASS	0.121	-
moonshotai/kimi-k3	0.703	$5.9242	13.8s	0.933	0.480	0.943	0.578	0.422	0.886	0.533	0.667	0.728	0.864	0.800	0.599	FAIL (1 FP)	PASS	0.089	-
google/gemini-2.5-pro	0.700	$4.1601	14.2s	0.933	0.400	0.943	0.490	0.829	0.750	0.733	0.800	0.658	0.633	0.800	0.432	FAIL (1 FP)	FAIL (1 FP)	0.051	-
mistralai/mistral-medium-3.1	0.699	$0.4847	0.7s	0.978	0.747	0.444	0.429	0.821	0.711	0.500	1.000	0.514	0.681	0.833	0.727	PASS	PASS	0.046	-
google/gemini-2.5-flash	0.694	$0.3796	0.8s	1.000	0.473	0.889	0.500	0.800	1.000	0.333	0.800	0.625	0.500	0.833	0.571	PASS	PASS	0.013	-
openai/gpt-oss-120b	0.692	$0.0634	6.5s	0.943	0.820	0.855	0.475	0.305	0.956	0.393	0.728	0.668	0.762	0.662	0.743	FAIL (1 FP)	FAIL (1 FP)	0.126	-
deepseek/deepseek-v4-pro	0.680	$0.6692	16.2s	0.851	0.280	0.876	0.581	0.532	0.857	0.867	0.700	0.764	0.597	0.800	0.456	FAIL (1 FP)	PASS	0.162	-
google/gemma-4-31b-it	0.667	$0.1181	2.4s	0.876	0.400	0.893	0.476	0.668	0.828	0.533	0.960	0.667	0.550	0.675	0.476	FAIL (1 FP)	PASS	0.137	-
openai/gpt-5.4	0.634	$2.7907	1.4s	0.818	0.420	0.878	0.653	0.706	0.565	0.333	0.674	0.589	0.670	0.785	0.520	FAIL (1 FP)	FAIL (3 FP)	0.091	-
minimax/minimax-m3	0.626	$0.3575	1.7s	0.507	0.533	0.547	0.429	0.798	0.853	0.800	0.780	0.552	0.599	0.399	0.719	FAIL (1 FP)	FAIL (1 FP)	0.201	-
google/gemini-2.5-flash-lite	0.623	$0.1132	0.8s	0.943	0.800	0.971	0.545	0.657	0.460	0.309	0.720	0.569	0.438	0.459	0.609	FAIL (2 FP)	PASS	0.070	-
qwen/qwen3.6-plus	0.607	$1.1534	36.9s	0.971	0.433	0.686	0.336	0.470	0.844	0.000	0.800	0.592	0.886	0.691	0.579	PASS	PASS	0.148	-
z-ai/glm-5.2	0.606	$0.7376	3.5s	0.910	0.320	0.813	0.508	0.543	0.555	0.547	0.603	0.601	0.527	0.678	0.670	FAIL (2 FP)	FAIL (5 FP)	0.117	-
deepseek/deepseek-v4-flash-0731	0.601	$0.1482	13.2s	0.905	0.667	0.706	0.295	0.365	0.743	0.400	0.740	0.612	0.943	0.636	0.203	PASS	PASS	0.187	-
deepseek/deepseek-v3.2	0.599	$0.3015	1.7s	0.857	0.633	0.667	0.616	0.859	0.613	0.473	0.667	0.303	0.500	0.476	0.527	PASS	PASS	0.072	-
meta/muse-spark-1.1	0.595	$2.2536	4.3s	0.813	0.433	0.771	0.456	0.604	0.766	0.267	0.667	0.623	0.680	0.604	0.451	PASS	PASS	0.224	-
openai/gpt-5.4-mini	0.582	$0.8470	1.1s	0.956	0.800	1.000	0.533	0.563	0.442	0.100	0.800	0.488	0.241	0.514	0.542	FAIL (2 FP)	FAIL (3 FP)	0.075	-
nvidia/nemotron-3-super-120b-a12b	0.577	$0.2194	22.3s	0.971	0.760	0.785	0.417	0.029	0.892	0.000	0.800	0.736	0.754	0.462	0.319	FAIL (1 FP)	PASS	0.111	-
deepseek/deepseek-r1	0.574	$1.4743	35.7s	0.857	0.360	0.848	0.352	0.365	0.746	0.267	0.800	0.543	0.462	0.620	0.669	FAIL (1 FP)	FAIL (3 FP)	0.140	-
google/gemma-4-26b-a4b-it	0.572	$0.0825	1.4s	0.950	0.920	0.428	0.170	0.412	0.867	0.133	0.667	0.714	0.697	0.545	0.358	FAIL (1 FP)	PASS	0.136	-
qwen/qwen3.5-27b	0.567	$0.8941	37.6s	0.373	0.640	0.598	0.417	0.625	0.800	0.600	0.693	0.751	0.586	0.373	0.350	PASS	PASS	0.214	-
mistralai/mistral-large-2512	0.547	$0.6184	3.3s	0.933	0.407	0.893	0.545	0.574	0.400	0.320	0.800	0.333	0.239	0.449	0.667	FAIL (1 FP)	PASS	0.043	-
deepseek/deepseek-r1-0528	0.540	$1.1645	29.4s	0.728	0.440	0.744	0.411	0.422	0.750	0.120	0.773	0.523	0.472	0.553	0.549	FAIL (1 FP)	FAIL (8 FP)	0.160	-
openai/gpt-oss-20b	0.540	$0.0620	9.2s	0.943	0.780	0.421	0.279	0.147	0.692	0.300	0.560	0.678	0.825	0.483	0.375	FAIL (1 FP)	FAIL (1 FP)	0.163	-
meta-llama/llama-4-maverick	0.522	$0.2206	1.1s	0.889	0.320	1.000	0.480	0.289	0.518	0.000	0.693	0.596	0.667	0.364	0.444	FAIL (1 FP)	PASS	0.033	-
qwen/qwen3.7-plus	0.506	$0.7238	23.7s	0.744	0.400	0.667	0.468	0.394	0.455	0.393	0.514	0.459	0.270	0.580	0.728	FAIL (2 FP)	FAIL (11 FP)	0.063	-
mistralai/codestral-2508	0.504	$0.3585	0.9s	0.800	0.400	0.800	0.307	0.452	0.371	0.400	0.610	0.487	0.300	0.401	0.727	PASS	FAIL (1 FP)	0.069	-
qwen/qwen3-235b-a22b-2507	0.499	$0.1078	2.5s	0.731	0.480	0.709	0.436	0.186	0.372	0.233	0.701	0.652	0.359	0.597	0.529	FAIL (1 FP)	FAIL (6 FP)	0.150	-
meta-llama/llama-3.3-70b-instruct	0.493	$0.1107	1.3s	0.800	0.567	0.774	0.295	0.040	0.659	0.267	0.480	0.769	0.571	0.329	0.363	PASS	PASS	0.117	-
tencent/hy3	0.489	$0.3876	22.3s	0.781	0.100	0.743	0.229	0.033	0.848	0.000	0.753	0.664	0.836	0.346	0.539	PASS	PASS	0.124	-
xiaomi/mimo-v2.5	0.471	$0.1692	4.2s	0.843	0.080	0.649	0.569	0.384	0.402	0.233	0.441	0.563	0.341	0.578	0.569	FAIL (2 FP)	FAIL (7 FP)	0.150	-
openai/o3	0.457	$3.3100	6.7s	0.507	0.000	0.545	0.356	0.464	0.857	0.333	0.733	0.635	0.451	0.425	0.178	PASS	PASS	0.188	-
nvidia/nemotron-3-ultra-550b-a55b	0.453	$0.7782	1.2s	0.813	0.473	0.547	0.512	0.091	0.598	0.333	0.513	0.380	0.244	0.463	0.471	FAIL (1 FP)	FAIL (4 FP)	0.231	-
meta-llama/llama-4-scout	0.442	$0.1111	0.9s	0.600	0.413	0.651	0.256	0.085	0.367	0.400	0.493	0.551	0.447	0.256	0.787	FAIL (1 FP)	PASS	0.224	-
xiaomi/mimo-v2.5-pro	0.438	$0.4954	2.2s	0.651	0.400	0.518	0.414	0.383	0.705	0.100	0.300	0.501	0.641	0.386	0.253	PASS	FAIL (1 FP)	0.217	-
cohere/command-a	0.425	$2.9415	1.7s	0.800	0.400	0.667	0.436	0.000	0.345	0.340	0.507	0.455	0.335	0.218	0.600	FAIL (3 FP)	FAIL (5 FP)	0.052	-
moonshotai/kimi-k2.6	0.401	$1.8854	55.7s	0.581	0.080	0.819	0.181	0.036	0.473	0.000	0.813	0.533	0.397	0.394	0.510	FAIL (1 FP)	FAIL (3 FP)	0.152	-
stepfun/step-3.7-flash	0.398	$0.4302	13.9s	0.886	0.733	0.240	0.000	-	0.781	0.000	0.633	0.366	-	0.279	0.067	PASS	PASS	0.156	15.2%
meituan/longcat-2.0	0.390	$0.5573	10.5s	0.000	0.233	0.613	0.436	0.400	0.781	0.333	0.500	0.482	0.280	0.152	0.473	PASS	PASS	0.164	-
deepseek/deepseek-r1-distill-llama-70b	0.387	$0.6913	27.2s	0.373	0.593	0.160	0.422	0.312	0.971	0.000	0.500	0.542	0.374	0.346	0.050	FAIL (1 FP)	PASS	0.109	-
qwen/qwen3-14b	0.385	$0.3989	14.4s	0.690	0.680	0.529	0.089	0.018	0.488	0.000	0.453	0.744	0.470	0.185	0.278	FAIL (1 FP)	FAIL (1 FP)	0.116	-
cohere/command-r-plus-08-2024	0.381	$2.8505	1.1s	0.613	0.267	0.133	0.361	0.139	0.712	0.000	0.600	0.751	0.460	0.239	0.295	PASS	PASS	0.148	-
thinkingmachines/inkling-small	0.358	$1.0333	18.5s	0.530	0.160	0.592	0.150	0.133	0.467	0.000	0.693	0.219	0.483	0.691	0.181	FAIL (1 FP)	FAIL (6 FP)	0.167	-
inclusionai/ring-2.6-1t	0.337	$0.2468	10.8s	0.507	0.867	0.522	0.057	0.031	0.537	0.000	0.400	0.547	0.350	0.094	0.129	PASS	PASS	0.140	-
openai/gpt-3.5-turbo	0.310	$0.5564	1.4s	0.803	0.400	0.222	0.622	0.055	0.263	0.000	0.444	0.286	0.211	0.200	0.218	FAIL (2 FP)	FAIL (11 FP)	0.021	-
microsoft/phi-4	0.299	$0.0792	0.5s	0.237	0.000	0.427	0.239	0.000	0.000	0.400	0.800	0.228	0.620	0.423	0.217	PASS	PASS	0.108	-
qwen/qwen3.8-max	0.272	$4.2714	29.7s	0.280	0.000	0.320	0.231	0.000	0.373	0.000	0.900	0.050	0.460	0.646	0.000	PASS	PASS	0.112	-
meta-llama/llama-3.1-8b-instruct	0.259	$0.0572	0.9s	0.200	0.593	0.274	0.133	0.167	0.203	0.700	0.263	0.124	0.090	0.140	0.216	PASS	FAIL (2 FP)	0.164	-
thinkingmachines/inkling	0.225	$3.2258	65.9s	0.307	0.000	0.427	0.124	0.042	0.200	0.000	0.627	0.315	0.460	0.202	0.000	PASS	FAIL (1 FP)	0.153	-
openai/o4-mini	0.135	$2.1179	7.3s	0.160	0.133	0.147	0.067	0.000	0.147	0.000	0.267	0.202	0.360	0.000	0.133	PASS	PASS	0.182	-
mistralai/mistral-small-2603	0.019	$0.1745	0.7s	0.000	0.000	0.000	0.000	0.000	0.000	0.000	0.000	0.000	0.000	0.000	0.222	PASS	PASS	0.000	-
qwen/qwen3-8b	0.000	$0.3256	39.8s	0.000	0.000	0.000	0.000	0.000	0.000	0.000	0.000	0.000	0.000	0.000	0.000	PASS	PASS	0.000	-

---

Detailed Results

Per-Model Detail

Full per-model profile: F1 averaged across episodes, total cost per episode at current pricing, p50 / p95 latency, JSON compliance, parse-failure rate, the distribution of extraction methods the parser had to use, and verbosity / truncation telemetry. The Extraction methods list shows how often each route was hit. json_array_direct is the cleanest; the rest are recovery paths. The verbosity row flags models that emit long reason fields or run out of token budget mid-response. Ordered by F1 descending so the best performers appear first.

claude-haiku-4-5-20251001

• F1 (avg across episodes): 0.842
• Total cost / episode: $1.0820
• p50 / p95 latency: 24.23s / 135.48s
• JSON compliance: 1.00
• JSON mode: native (100% native, 855 calls)
• Parse failure rate: 0.0%
• Extraction methods: json_array_direct: 855
• Verbosity: 0/855 calls over 1024 output tokens (0.0%); 0 hit max_tokens (0.0%); 0 salvaged from truncated JSON (0.0%)
• Segment category named on 716/716 detections (100%); the rest stay uncategorized (resolver: production)

claude-sonnet-4-6

• F1 (avg across episodes): 0.799
• Total cost / episode: $3.2386
• p50 / p95 latency: 4.11s / 83.49s
• JSON compliance: 1.00
• JSON mode: native (100% native, 855 calls)
• Parse failure rate: 0.0%
• Extraction methods: json_array_direct: 855
• Verbosity: 0/855 calls over 1024 output tokens (0.0%); 0 hit max_tokens (0.0%); 0 salvaged from truncated JSON (0.0%)
• Segment category named on 516/516 detections (100%); the rest stay uncategorized (resolver: production)

x-ai/grok-4.5

• F1 (avg across episodes): 0.797
• Total cost / episode: $3.3554
• p50 / p95 latency: 10.86s / 52.43s
• JSON compliance: 1.00
• JSON mode: native (100% native, 855 calls)
• Parse failure rate: 0.0%
• Extraction methods: json_array_direct: 855
• Verbosity: 311/855 calls over 1024 output tokens (36.4%); 0 hit max_tokens (0.0%); 0 salvaged from truncated JSON (0.0%)
• Segment category named on 451/451 detections (100%); the rest stay uncategorized (resolver: production)

google/gemini-3.6-flash

• F1 (avg across episodes): 0.796
• Total cost / episode: $2.8863
• p50 / p95 latency: 4.51s / 10.24s
• JSON compliance: 1.00
• JSON mode: native (100% native, 855 calls)
• Parse failure rate: 0.0%
• Extraction methods: json_array_direct: 854, json_object_single_ad_truncated: 1
• Verbosity: 302/855 calls over 1024 output tokens (35.3%); 1 hit max_tokens (0.1%); 1 salvaged from truncated JSON (0.1%)
• Segment category named on 414/415 detections (100%); the rest stay uncategorized (resolver: production)

x-ai/grok-4.3

• F1 (avg across episodes): 0.789
• Total cost / episode: $1.6048
• p50 / p95 latency: 3.73s / 10.34s
• JSON compliance: 1.00
• JSON mode: native (100% native, 855 calls)
• Parse failure rate: 0.0%
• Extraction methods: json_array_direct: 855
• Verbosity: 115/855 calls over 1024 output tokens (13.5%); 0 hit max_tokens (0.0%); 0 salvaged from truncated JSON (0.0%)
• Segment category named on 684/684 detections (100%); the rest stay uncategorized (resolver: production)

google/gemini-3.5-flash

• F1 (avg across episodes): 0.787
• Total cost / episode: $3.4202
• p50 / p95 latency: 5.59s / 11.20s
• JSON compliance: 1.00
• JSON mode: native (100% native, 855 calls)
• Parse failure rate: 0.0%
• Extraction methods: json_array_direct: 852, regex_json_array: 3
• Verbosity: 418/855 calls over 1024 output tokens (48.9%); 0 hit max_tokens (0.0%); 0 salvaged from truncated JSON (0.0%)
• Segment category named on 439/439 detections (100%); the rest stay uncategorized (resolver: production)

mistralai/mistral-medium-3-5

• F1 (avg across episodes): 0.787
• Total cost / episode: $1.8393
• p50 / p95 latency: 1.10s / 3.29s
• JSON compliance: 1.00
• JSON mode: native (100% native, 855 calls)
• Parse failure rate: 0.0%
• Extraction methods: json_array_direct: 855
• Verbosity: 1/855 calls over 1024 output tokens (0.1%); 0 hit max_tokens (0.0%); 0 salvaged from truncated JSON (0.0%)
• Segment category named on 833/833 detections (100%); the rest stay uncategorized (resolver: production)

openai/gpt-5.5

• F1 (avg across episodes): 0.784
• Total cost / episode: $7.6842
• p50 / p95 latency: 5.36s / 23.68s
• JSON compliance: 0.87
• JSON mode: native (100% native, 855 calls)
• Parse failure rate: 0.4%
• Extraction methods: json_object_no_ads: 489, json_object_single_ad: 363, parse_failure: 3
• Verbosity: 108/855 calls over 1024 output tokens (12.6%); 3 hit max_tokens (0.4%); 0 salvaged from truncated JSON (0.0%)
• Segment category named on 363/363 detections (100%); the rest stay uncategorized (resolver: production)

claude-fable-5

• F1 (avg across episodes): 0.783
• Total cost / episode: $10.7552
• p50 / p95 latency: 6.61s / 68.22s
• JSON compliance: 1.00
• JSON mode: native (100% native, 855 calls)
• Parse failure rate: 0.0%
• Extraction methods: json_array_direct: 855
• Verbosity: 0/855 calls over 1024 output tokens (0.0%); 0 hit max_tokens (0.0%); 0 salvaged from truncated JSON (0.0%)
• Segment category named on 566/566 detections (100%); the rest stay uncategorized (resolver: production)

qwen/qwen3.5-plus-02-15

• F1 (avg across episodes): 0.779
• Total cost / episode: $1.0746
• p50 / p95 latency: 29.18s / 52.05s
• JSON compliance: 1.00
• JSON mode: native (100% native, 855 calls)
• Parse failure rate: 0.5%
• Extraction methods: json_array_direct: 851, parse_failure: 4
• Verbosity: 809/855 calls over 1024 output tokens (94.6%); 0 hit max_tokens (0.0%); 0 salvaged from truncated JSON (0.0%)
• Segment category named on 700/700 detections (100%); the rest stay uncategorized (resolver: production)

google/gemini-3.5-flash-lite

• F1 (avg across episodes): 0.776
• Total cost / episode: $0.3643
• p50 / p95 latency: 0.67s / 1.42s
• JSON compliance: 1.00
• JSON mode: native (100% native, 855 calls)
• Parse failure rate: 0.0%
• Extraction methods: json_array_direct: 853, json_object_single_ad_truncated: 1, regex_json_array: 1
• Verbosity: 0/855 calls over 1024 output tokens (0.0%); 0 hit max_tokens (0.0%); 1 salvaged from truncated JSON (0.1%)
• Segment category named on 664/665 detections (100%); the rest stay uncategorized (resolver: production)

openai/gpt-5.6-terra

• F1 (avg across episodes): 0.768
• Total cost / episode: $1.1951
• p50 / p95 latency: 2.10s / 6.71s
• JSON compliance: 0.86
• JSON mode: native (100% native, 855 calls)
• Parse failure rate: 0.0%
• Extraction methods: json_object_no_ads: 458, json_object_single_ad: 397
• Verbosity: 1/855 calls over 1024 output tokens (0.1%); 0 hit max_tokens (0.0%); 0 salvaged from truncated JSON (0.0%)
• Segment category named on 365/365 detections (100%); the rest stay uncategorized (resolver: production)

claude-opus-4-8

• F1 (avg across episodes): 0.760
• Total cost / episode: $5.3661
• p50 / p95 latency: 7.76s / 68.92s
• JSON compliance: 1.00
• JSON mode: native (100% native, 855 calls)
• Parse failure rate: 0.0%
• Extraction methods: json_array_direct: 855
• Verbosity: 0/855 calls over 1024 output tokens (0.0%); 0 hit max_tokens (0.0%); 0 salvaged from truncated JSON (0.0%)
• Segment category named on 605/605 detections (100%); the rest stay uncategorized (resolver: production)

openai/gpt-5.6-luna

• F1 (avg across episodes): 0.754
• Total cost / episode: $0.1402
• p50 / p95 latency: 3.46s / 11.35s
• JSON compliance: 0.83
• JSON mode: native (100% native, 855 calls)
• Parse failure rate: 1.1%
• Extraction methods: json_object_ads_key: 32, json_object_no_ads: 368, json_object_single_ad: 446, parse_failure: 9
• Verbosity: 45/855 calls over 1024 output tokens (5.3%); 9 hit max_tokens (1.1%); 0 salvaged from truncated JSON (0.0%)
• Segment category named on 390/390 detections (100%); the rest stay uncategorized (resolver: production)

google/gemini-3.1-pro-preview

• F1 (avg across episodes): 0.752
• Total cost / episode: $4.9160
• p50 / p95 latency: 8.35s / 20.73s
• JSON compliance: 1.00
• JSON mode: native (100% native, 855 calls)
• Parse failure rate: 0.0%
• Extraction methods: json_array_direct: 855
• Verbosity: 397/855 calls over 1024 output tokens (46.4%); 0 hit max_tokens (0.0%); 0 salvaged from truncated JSON (0.0%)
• Segment category named on 463/463 detections (100%); the rest stay uncategorized (resolver: production)

claude-opus-5

• F1 (avg across episodes): 0.748
• Total cost / episode: $5.3761
• p50 / p95 latency: 3.67s / 64.55s
• JSON compliance: 1.00
• JSON mode: native (100% native, 855 calls)
• Parse failure rate: 0.0%
• Extraction methods: json_array_direct: 855
• Verbosity: 0/855 calls over 1024 output tokens (0.0%); 0 hit max_tokens (0.0%); 0 salvaged from truncated JSON (0.0%)
• Segment category named on 547/547 detections (100%); the rest stay uncategorized (resolver: production)

google/gemini-3.1-flash-lite

• F1 (avg across episodes): 0.747
• Total cost / episode: $0.2988
• p50 / p95 latency: 0.85s / 2.11s
• JSON compliance: 0.94
• JSON mode: native (100% native, 855 calls)
• Parse failure rate: 0.0%
• Extraction methods: json_array_direct: 771, json_object_single_ad_truncated: 1, regex_json_array: 83
• Verbosity: 0/855 calls over 1024 output tokens (0.0%); 0 hit max_tokens (0.0%); 1 salvaged from truncated JSON (0.1%)
• Segment category named on 841/842 detections (100%); the rest stay uncategorized (resolver: production)

qwen/qwen3.6-flash

• F1 (avg across episodes): 0.747
• Total cost / episode: $0.5435
• p50 / p95 latency: 7.55s / 19.48s
• JSON compliance: 0.96
• JSON mode: native (100% native, 855 calls)
• Parse failure rate: 3.5%
• Extraction methods: json_array_direct: 820, json_object_single_ad_truncated: 5, parse_failure: 30
• Verbosity: 568/855 calls over 1024 output tokens (66.4%); 35 hit max_tokens (4.1%); 5 salvaged from truncated JSON (0.6%)
• Segment category named on 415/420 detections (99%); the rest stay uncategorized (resolver: production)

qwen/qwen3.7-max

• F1 (avg across episodes): 0.746
• Total cost / episode: $2.8109
• p50 / p95 latency: 23.41s / 62.65s
• JSON compliance: 0.99
• JSON mode: native (100% native, 855 calls)
• Parse failure rate: 0.5%
• Extraction methods: json_array_direct: 833, json_object_single_ad: 8, json_object_single_ad_truncated: 10, parse_failure: 4
• Verbosity: 507/855 calls over 1024 output tokens (59.3%); 14 hit max_tokens (1.6%); 10 salvaged from truncated JSON (1.2%)
• Segment category named on 752/762 detections (99%); the rest stay uncategorized (resolver: production)

openai/gpt-5.6-sol

• F1 (avg across episodes): 0.738
• Total cost / episode: $6.2293
• p50 / p95 latency: 3.72s / 11.86s
• JSON compliance: 0.84
• JSON mode: native (100% native, 855 calls)
• Parse failure rate: 0.0%
• Extraction methods: json_object_ads_key: 53, json_object_no_ads: 368, json_object_segments_key: 3, json_object_single_ad: 431
• Verbosity: 9/855 calls over 1024 output tokens (1.1%); 0 hit max_tokens (0.0%); 0 salvaged from truncated JSON (0.0%)
• Segment category named on 411/412 detections (100%); the rest stay uncategorized (resolver: production)

claude-sonnet-5

• F1 (avg across episodes): 0.732
• Total cost / episode: $2.1452
• p50 / p95 latency: 8.48s / 66.02s
• JSON compliance: 1.00
• JSON mode: native (100% native, 855 calls)
• Parse failure rate: 0.0%
• Extraction methods: json_array_direct: 854, json_object_single_ad: 1
• Verbosity: 0/855 calls over 1024 output tokens (0.0%); 0 hit max_tokens (0.0%); 0 salvaged from truncated JSON (0.0%)
• Segment category named on 612/612 detections (100%); the rest stay uncategorized (resolver: production)

claude-opus-4-7

• F1 (avg across episodes): 0.731
• Total cost / episode: $5.3293
• p50 / p95 latency: 4.60s / 66.02s
• JSON compliance: 1.00
• JSON mode: native (100% native, 855 calls)
• Parse failure rate: 0.0%
• Extraction methods: json_array_direct: 855
• Verbosity: 0/855 calls over 1024 output tokens (0.0%); 0 hit max_tokens (0.0%); 0 salvaged from truncated JSON (0.0%)
• Segment category named on 615/615 detections (100%); the rest stay uncategorized (resolver: production)

deepseek/deepseek-v4-flash

• F1 (avg across episodes): 0.708
• Total cost / episode: $0.2046
• p50 / p95 latency: 6.27s / 40.26s
• JSON compliance: 0.82
• JSON mode: native (100% native, 855 calls)
• Parse failure rate: 5.5%
• Extraction methods: json_array_direct: 93, json_object_ads_key: 175, json_object_no_ads: 266, json_object_segments_key: 6, json_object_single_ad: 268, parse_failure: 47
• Verbosity: 299/855 calls over 1024 output tokens (35.0%); 46 hit max_tokens (5.4%); 0 salvaged from truncated JSON (0.0%)
• Segment category named on 362/362 detections (100%); the rest stay uncategorized (resolver: production)

qwen/qwen3.7-flash

• F1 (avg across episodes): 0.706
• Total cost / episode: $0.0725
• p50 / p95 latency: 10.14s / 27.58s
• JSON compliance: 0.96
• JSON mode: native (100% native, 855 calls)
• Parse failure rate: 3.9%
• Extraction methods: json_array_direct: 810, json_object_single_ad_truncated: 12, parse_failure: 33
• Verbosity: 559/855 calls over 1024 output tokens (65.4%); 44 hit max_tokens (5.1%); 12 salvaged from truncated JSON (1.4%)
• Segment category named on 377/388 detections (97%); the rest stay uncategorized (resolver: production)

moonshotai/kimi-k3

• F1 (avg across episodes): 0.703
• Total cost / episode: $5.9242
• p50 / p95 latency: 13.78s / 96.34s
• JSON compliance: 0.81
• JSON mode: native (100% native, 855 calls)
• Parse failure rate: 5.8%
• Extraction methods: json_array_direct: 361, json_object_ads_key: 181, json_object_no_ads: 27, json_object_single_ad: 6, json_object_single_ad_truncated: 122, parse_failure: 50, regex_json_array: 108
• Verbosity: 276/855 calls over 1024 output tokens (32.3%); 53 hit max_tokens (6.2%); 122 salvaged from truncated JSON (14.3%)
• Segment category named on 264/386 detections (68%); the rest stay uncategorized (resolver: production)

google/gemini-2.5-pro

• F1 (avg across episodes): 0.700
• Total cost / episode: $4.1601
• p50 / p95 latency: 14.22s / 34.09s
• JSON compliance: 0.96
• JSON mode: native (100% native, 855 calls)
• Parse failure rate: 1.4%
• Extraction methods: json_array_direct: 805, json_object_single_ad_truncated: 9, parse_failure: 12, regex_json_array: 29
• Verbosity: 637/855 calls over 1024 output tokens (74.5%); 9 hit max_tokens (1.1%); 9 salvaged from truncated JSON (1.1%)
• Segment category named on 504/512 detections (98%); the rest stay uncategorized (resolver: production)

mistralai/mistral-medium-3.1

• F1 (avg across episodes): 0.699
• Total cost / episode: $0.4847
• p50 / p95 latency: 0.74s / 7.65s
• JSON compliance: 1.00
• JSON mode: native (100% native, 855 calls)
• Parse failure rate: 0.0%
• Extraction methods: json_array_direct: 855
• Verbosity: 0/855 calls over 1024 output tokens (0.0%); 0 hit max_tokens (0.0%); 0 salvaged from truncated JSON (0.0%)
• Segment category named on 705/705 detections (100%); the rest stay uncategorized (resolver: production)

google/gemini-2.5-flash

• F1 (avg across episodes): 0.694
• Total cost / episode: $0.3796
• p50 / p95 latency: 0.83s / 2.21s
• JSON compliance: 1.00
• JSON mode: native (100% native, 855 calls)
• Parse failure rate: 0.0%
• Extraction methods: json_array_direct: 852, json_object_single_ad_truncated: 3
• Verbosity: 0/855 calls over 1024 output tokens (0.0%); 0 hit max_tokens (0.0%); 3 salvaged from truncated JSON (0.4%)
• Segment category named on 973/975 detections (100%); the rest stay uncategorized (resolver: production)

openai/gpt-oss-120b

• F1 (avg across episodes): 0.692
• Total cost / episode: $0.0634
• p50 / p95 latency: 6.54s / 44.55s
• JSON compliance: 0.88
• JSON mode: native (100% native, 855 calls)
• Parse failure rate: 1.1%
• Extraction methods: json_array_direct: 28, json_object_ads_key: 181, json_object_no_ads: 429, json_object_single_ad: 202, json_object_single_ad_truncated: 1, parse_failure: 9, regex_json_array: 5
• Verbosity: 224/855 calls over 1024 output tokens (26.2%); 3 hit max_tokens (0.4%); 1 salvaged from truncated JSON (0.1%)
• Segment category named on 700/701 detections (100%); the rest stay uncategorized (resolver: production)

deepseek/deepseek-v4-pro

• F1 (avg across episodes): 0.680
• Total cost / episode: $0.6692
• p50 / p95 latency: 16.17s / 68.15s
• JSON compliance: 0.87
• JSON mode: native (100% native, 855 calls)
• Parse failure rate: 5.3%
• Extraction methods: json_array_direct: 434, json_object_ads_key: 10, json_object_no_ads: 39, json_object_segments_key: 205, json_object_single_ad: 118, json_object_single_ad_truncated: 3, parse_failure: 45, regex_json_array: 1
• Verbosity: 376/855 calls over 1024 output tokens (44.0%); 39 hit max_tokens (4.6%); 3 salvaged from truncated JSON (0.4%)
• Segment category named on 282/288 detections (98%); the rest stay uncategorized (resolver: production)

google/gemma-4-31b-it

• F1 (avg across episodes): 0.667
• Total cost / episode: $0.1181
• p50 / p95 latency: 2.45s / 11.72s
• JSON compliance: 0.87
• JSON mode: native (100% native, 855 calls)
• Parse failure rate: 0.0%
• Extraction methods: json_array_direct: 14, json_object_ads_key: 378, json_object_no_ads: 283, json_object_single_ad: 180
• Verbosity: 4/855 calls over 1024 output tokens (0.5%); 0 hit max_tokens (0.0%); 0 salvaged from truncated JSON (0.0%)
• Segment category named on 569/569 detections (100%); the rest stay uncategorized (resolver: production)

openai/gpt-5.4

• F1 (avg across episodes): 0.634
• Total cost / episode: $2.7907
• p50 / p95 latency: 1.43s / 2.57s
• JSON compliance: 0.80
• JSON mode: native (100% native, 855 calls)
• Parse failure rate: 0.0%
• Extraction methods: json_object_no_ads: 274, json_object_single_ad: 581
• Verbosity: 0/855 calls over 1024 output tokens (0.0%); 0 hit max_tokens (0.0%); 0 salvaged from truncated JSON (0.0%)
• Segment category named on 470/470 detections (100%); the rest stay uncategorized (resolver: production)

minimax/minimax-m3

• F1 (avg across episodes): 0.626
• Total cost / episode: $0.3575
• p50 / p95 latency: 1.68s / 6.67s
• JSON compliance: 0.88
• JSON mode: native (100% native, 855 calls)
• Parse failure rate: 0.5%
• Extraction methods: json_array_direct: 340, json_object_ads_key: 4, json_object_no_ads: 262, json_object_single_ad: 48, markdown_code_block: 188, parse_failure: 4, regex_json_array: 9
• Verbosity: 18/855 calls over 1024 output tokens (2.1%); 2 hit max_tokens (0.2%); 0 salvaged from truncated JSON (0.0%)
• Segment category named on 470/471 detections (100%); the rest stay uncategorized (resolver: production)

google/gemini-2.5-flash-lite

• F1 (avg across episodes): 0.623
• Total cost / episode: $0.1132
• p50 / p95 latency: 0.84s / 1.75s
• JSON compliance: 0.97
• JSON mode: native (100% native, 855 calls)
• Parse failure rate: 2.0%
• Extraction methods: json_array_direct: 791, json_object_single_ad_truncated: 47, parse_failure: 17
• Verbosity: 6/855 calls over 1024 output tokens (0.7%); 0 hit max_tokens (0.0%); 47 salvaged from truncated JSON (5.5%)
• Segment category named on 945/987 detections (96%); the rest stay uncategorized (resolver: production)

qwen/qwen3.6-plus

• F1 (avg across episodes): 0.607
• Total cost / episode: $1.1534
• p50 / p95 latency: 36.91s / 70.77s
• JSON compliance: 0.90
• JSON mode: native (100% native, 855 calls)
• Parse failure rate: 10.1%
• Extraction methods: json_array_direct: 748, json_object_single_ad_truncated: 21, parse_failure: 86
• Verbosity: 789/855 calls over 1024 output tokens (92.3%); 107 hit max_tokens (12.5%); 21 salvaged from truncated JSON (2.5%)
• Segment category named on 444/464 detections (96%); the rest stay uncategorized (resolver: production)

z-ai/glm-5.2

• F1 (avg across episodes): 0.606
• Total cost / episode: $0.7376
• p50 / p95 latency: 3.55s / 24.22s
• JSON compliance: 0.73
• JSON mode: native (100% native, 855 calls)
• Parse failure rate: 3.4%
• Extraction methods: json_array_direct: 5, json_object_ad_key: 4, json_object_ads_key: 95, json_object_no_ads: 86, json_object_single_ad: 635, json_object_single_ad_truncated: 1, parse_failure: 29
• Verbosity: 103/855 calls over 1024 output tokens (12.0%); 6 hit max_tokens (0.7%); 1 salvaged from truncated JSON (0.1%)
• Segment category named on 522/529 detections (99%); the rest stay uncategorized (resolver: production)

deepseek/deepseek-v4-flash-0731

• F1 (avg across episodes): 0.601
• Total cost / episode: $0.1482
• p50 / p95 latency: 13.20s / 61.85s
• JSON compliance: 0.73
• JSON mode: native (100% native, 855 calls)
• Parse failure rate: 18.5%
• Extraction methods: json_array_direct: 224, json_object_ads_key: 9, json_object_no_ads: 222, json_object_single_ad: 238, json_object_single_ad_truncated: 3, parse_failure: 158, regex_json_array: 1
• Verbosity: 394/855 calls over 1024 output tokens (46.1%); 155 hit max_tokens (18.1%); 3 salvaged from truncated JSON (0.4%)
• Segment category named on 215/218 detections (99%); the rest stay uncategorized (resolver: production)

deepseek/deepseek-v3.2

• F1 (avg across episodes): 0.599
• Total cost / episode: $0.3015
• p50 / p95 latency: 1.74s / 10.84s
• JSON compliance: 1.00
• JSON mode: native (100% native, 855 calls)
• Parse failure rate: 0.0%
• Extraction methods: json_array_direct: 853, json_object_ads_key: 2
• Verbosity: 0/855 calls over 1024 output tokens (0.0%); 0 hit max_tokens (0.0%); 0 salvaged from truncated JSON (0.0%)
• Segment category named on 432/432 detections (100%); the rest stay uncategorized (resolver: production)

meta/muse-spark-1.1

• F1 (avg across episodes): 0.595
• Total cost / episode: $2.2536
• p50 / p95 latency: 4.31s / 13.59s
• JSON compliance: 0.93
• JSON mode: native (100% native, 855 calls)
• Parse failure rate: 0.5%
• Extraction methods: json_object_ads_key: 28, json_object_no_ads: 643, json_object_segments_key: 3, json_object_single_ad: 175, json_object_single_ad_truncated: 2, parse_failure: 4
• Verbosity: 432/855 calls over 1024 output tokens (50.5%); 6 hit max_tokens (0.7%); 2 salvaged from truncated JSON (0.2%)
• Segment category named on 235/236 detections (100%); the rest stay uncategorized (resolver: production)

openai/gpt-5.4-mini

• F1 (avg across episodes): 0.582
• Total cost / episode: $0.8470
• p50 / p95 latency: 1.13s / 2.45s
• JSON compliance: 0.78
• JSON mode: native (100% native, 855 calls)
• Parse failure rate: 0.0%
• Extraction methods: json_object_no_ads: 218, json_object_single_ad: 637
• Verbosity: 0/855 calls over 1024 output tokens (0.0%); 0 hit max_tokens (0.0%); 0 salvaged from truncated JSON (0.0%)
• Segment category named on 599/599 detections (100%); the rest stay uncategorized (resolver: production)

nvidia/nemotron-3-super-120b-a12b

• F1 (avg across episodes): 0.577
• Total cost / episode: $0.2194
• p50 / p95 latency: 22.33s / 248.36s
• JSON compliance: 0.79
• JSON mode: native (100% native, 855 calls)
• Parse failure rate: 14.9%
• Extraction methods: json_array_direct: 520, json_object_ads_key: 1, json_object_no_ads: 40, json_object_single_ad: 160, json_object_single_ad_truncated: 6, parse_failure: 127, regex_json_array: 1
• Verbosity: 486/855 calls over 1024 output tokens (56.8%); 128 hit max_tokens (15.0%); 6 salvaged from truncated JSON (0.7%)
• Segment category named on 284/290 detections (98%); the rest stay uncategorized (resolver: production)

deepseek/deepseek-r1

• F1 (avg across episodes): 0.574
• Total cost / episode: $1.4743
• p50 / p95 latency: 35.66s / 195.41s
• JSON compliance: 0.84
• JSON mode: native (100% native, 855 calls)
• Parse failure rate: 9.9%
• Extraction methods: json_array_direct: 494, json_object_ads_key: 182, json_object_segments_key: 1, json_object_single_ad: 90, json_object_single_ad_truncated: 3, parse_failure: 85
• Verbosity: 387/855 calls over 1024 output tokens (45.3%); 88 hit max_tokens (10.3%); 3 salvaged from truncated JSON (0.4%)
• Segment category named on 654/657 detections (100%); the rest stay uncategorized (resolver: production)

google/gemma-4-26b-a4b-it

• F1 (avg across episodes): 0.572
• Total cost / episode: $0.0825
• p50 / p95 latency: 1.40s / 6.52s
• JSON compliance: 0.84
• JSON mode: native (100% native, 855 calls)
• Parse failure rate: 0.0%
• Extraction methods: json_object_ads_key: 73, json_object_no_ads: 372, json_object_single_ad: 410
• Verbosity: 0/855 calls over 1024 output tokens (0.0%); 0 hit max_tokens (0.0%); 0 salvaged from truncated JSON (0.0%)
• Segment category named on 495/495 detections (100%); the rest stay uncategorized (resolver: production)

qwen/qwen3.5-27b

• F1 (avg across episodes): 0.567
• Total cost / episode: $0.8941
• p50 / p95 latency: 37.62s / 126.80s
• JSON compliance: 0.70
• JSON mode: native (100% native, 855 calls)
• Parse failure rate: 29.6%
• Extraction methods: json_array_direct: 460, json_object_no_ads: 124, json_object_single_ad: 17, json_object_single_ad_truncated: 1, parse_failure: 253
• Verbosity: 574/855 calls over 1024 output tokens (67.1%); 132 hit max_tokens (15.4%); 1 salvaged from truncated JSON (0.1%)
• Segment category named on 272/273 detections (100%); the rest stay uncategorized (resolver: production)

mistralai/mistral-large-2512

• F1 (avg across episodes): 0.547
• Total cost / episode: $0.6184
• p50 / p95 latency: 3.28s / 7.58s
• JSON compliance: 1.00
• JSON mode: native (100% native, 855 calls)
• Parse failure rate: 0.0%
• Extraction methods: json_array_direct: 855
• Verbosity: 0/855 calls over 1024 output tokens (0.0%); 0 hit max_tokens (0.0%); 0 salvaged from truncated JSON (0.0%)
• Segment category named on 1280/1280 detections (100%); the rest stay uncategorized (resolver: production)

deepseek/deepseek-r1-0528

• F1 (avg across episodes): 0.540
• Total cost / episode: $1.1645
• p50 / p95 latency: 29.36s / 208.22s
• JSON compliance: 0.84
• JSON mode: native (100% native, 855 calls)
• Parse failure rate: 11.2%
• Extraction methods: json_array_direct: 535, json_object_ads_key: 154, json_object_no_ads: 7, json_object_segments_key: 4, json_object_single_ad: 55, json_object_single_ad_truncated: 4, parse_failure: 96
• Verbosity: 391/855 calls over 1024 output tokens (45.7%); 100 hit max_tokens (11.7%); 4 salvaged from truncated JSON (0.5%)
• Segment category named on 644/650 detections (99%); the rest stay uncategorized (resolver: production)

openai/gpt-oss-20b

• F1 (avg across episodes): 0.540
• Total cost / episode: $0.0620
• p50 / p95 latency: 9.18s / 51.77s
• JSON compliance: 0.80
• JSON mode: native (100% native, 855 calls)
• Parse failure rate: 11.6%
• Extraction methods: json_array_direct: 26, json_object_ads_key: 121, json_object_no_ads: 430, json_object_single_ad: 176, json_object_single_ad_truncated: 3, parse_failure: 99
• Verbosity: 349/855 calls over 1024 output tokens (40.8%); 94 hit max_tokens (11.0%); 3 salvaged from truncated JSON (0.4%)
• Segment category named on 486/490 detections (99%); the rest stay uncategorized (resolver: production)

meta-llama/llama-4-maverick

• F1 (avg across episodes): 0.522
• Total cost / episode: $0.2206
• p50 / p95 latency: 1.12s / 3.59s
• JSON compliance: 0.79
• JSON mode: native (100% native, 855 calls)
• Parse failure rate: 0.0%
• Extraction methods: json_object_no_ads: 259, json_object_single_ad: 596
• Verbosity: 1/855 calls over 1024 output tokens (0.1%); 0 hit max_tokens (0.0%); 0 salvaged from truncated JSON (0.0%)
• Segment category named on 476/476 detections (100%); the rest stay uncategorized (resolver: production)

qwen/qwen3.7-plus

• F1 (avg across episodes): 0.506
• Total cost / episode: $0.7238
• p50 / p95 latency: 23.68s / 62.94s
• JSON compliance: 0.80
• JSON mode: native (100% native, 855 calls)
• Parse failure rate: 2.6%
• Extraction methods: json_array_direct: 242, json_object_ads_key: 3, json_object_no_ads: 86, json_object_single_ad: 500, json_object_single_ad_truncated: 2, parse_failure: 22
• Verbosity: 678/855 calls over 1024 output tokens (79.3%); 23 hit max_tokens (2.7%); 2 salvaged from truncated JSON (0.2%)
• Segment category named on 859/863 detections (100%); the rest stay uncategorized (resolver: production)

mistralai/codestral-2508

• F1 (avg across episodes): 0.504
• Total cost / episode: $0.3585
• p50 / p95 latency: 0.88s / 2.54s
• JSON compliance: 1.00
• JSON mode: native (100% native, 855 calls)
• Parse failure rate: 0.0%
• Extraction methods: json_array_direct: 855
• Verbosity: 0/855 calls over 1024 output tokens (0.0%); 0 hit max_tokens (0.0%); 0 salvaged from truncated JSON (0.0%)
• Segment category named on 820/820 detections (100%); the rest stay uncategorized (resolver: production)

qwen/qwen3-235b-a22b-2507

• F1 (avg across episodes): 0.499
• Total cost / episode: $0.1078
• p50 / p95 latency: 2.51s / 7.43s
• JSON compliance: 0.84
• JSON mode: native (100% native, 855 calls)
• Parse failure rate: 0.0%
• Extraction methods: json_array_direct: 227, json_object_no_ads: 169, json_object_single_ad: 459
• Verbosity: 0/855 calls over 1024 output tokens (0.0%); 0 hit max_tokens (0.0%); 0 salvaged from truncated JSON (0.0%)
• Segment category named on 633/633 detections (100%); the rest stay uncategorized (resolver: production)

meta-llama/llama-3.3-70b-instruct

• F1 (avg across episodes): 0.493
• Total cost / episode: $0.1107
• p50 / p95 latency: 1.29s / 8.60s
• JSON compliance: 0.88
• JSON mode: native (100% native, 855 calls)
• Parse failure rate: 0.0%
• Extraction methods: json_array_direct: 147, json_object_no_ads: 380, json_object_single_ad: 322, json_object_single_ad_truncated: 4, regex_json_array: 2
• Verbosity: 0/855 calls over 1024 output tokens (0.0%); 0 hit max_tokens (0.0%); 4 salvaged from truncated JSON (0.5%)
• Segment category named on 466/470 detections (99%); the rest stay uncategorized (resolver: production)

tencent/hy3

• F1 (avg across episodes): 0.489
• Total cost / episode: $0.3876
• p50 / p95 latency: 22.27s / 59.98s
• JSON compliance: 0.62
• JSON mode: native (100% native, 855 calls)
• Parse failure rate: 27.6%
• Extraction methods: bracket_fallback: 12, json_array_direct: 144, json_object_ads_key: 10, json_object_no_ads: 198, json_object_single_ad: 249, json_object_single_ad_truncated: 6, parse_failure: 236
• Verbosity: 786/855 calls over 1024 output tokens (91.9%); 239 hit max_tokens (28.0%); 6 salvaged from truncated JSON (0.7%)
• Segment category named on 165/171 detections (96%); the rest stay uncategorized (resolver: production)

xiaomi/mimo-v2.5

• F1 (avg across episodes): 0.471
• Total cost / episode: $0.1692
• p50 / p95 latency: 4.16s / 21.43s
• JSON compliance: 0.73
• JSON mode: native (100% native, 855 calls)
• Parse failure rate: 3.0%
• Extraction methods: json_object_ads_detected_key: 1, json_object_ads_key: 54, json_object_no_ads: 160, json_object_segments_key: 2, json_object_single_ad: 492, json_object_single_ad_truncated: 48, markdown_code_block: 18, parse_failure: 26, regex_json_array: 54
• Verbosity: 36/855 calls over 1024 output tokens (4.2%); 20 hit max_tokens (2.3%); 48 salvaged from truncated JSON (5.6%)
• Segment category named on 533/587 detections (91%); the rest stay uncategorized (resolver: production)

openai/o3

• F1 (avg across episodes): 0.457
• Total cost / episode: $3.3100
• p50 / p95 latency: 6.66s / 20.44s
• JSON compliance: 0.93
• JSON mode: native (100% native, 855 calls)
• Parse failure rate: 0.7%
• Extraction methods: json_object_ads_key: 87, json_object_no_ads: 613, json_object_segments_key: 30, json_object_single_ad: 119, parse_failure: 6
• Verbosity: 280/855 calls over 1024 output tokens (32.7%); 0 hit max_tokens (0.0%); 0 salvaged from truncated JSON (0.0%)
• Segment category named on 140/140 detections (100%); the rest stay uncategorized (resolver: production)

nvidia/nemotron-3-ultra-550b-a55b

• F1 (avg across episodes): 0.453
• Total cost / episode: $0.7782
• p50 / p95 latency: 1.24s / 5.94s
• JSON compliance: 0.89
• JSON mode: native (100% native, 855 calls)
• Parse failure rate: 0.4%
• Extraction methods: json_array_direct: 448, json_object_ads_key: 59, json_object_no_ads: 56, json_object_segments_key: 37, json_object_single_ad: 252, parse_failure: 3
• Verbosity: 25/855 calls over 1024 output tokens (2.9%); 1 hit max_tokens (0.1%); 0 salvaged from truncated JSON (0.0%)
• Segment category named on 229/229 detections (100%); the rest stay uncategorized (resolver: production)

meta-llama/llama-4-scout

• F1 (avg across episodes): 0.442
• Total cost / episode: $0.1111
• p50 / p95 latency: 0.93s / 4.49s
• JSON compliance: 0.83
• JSON mode: native (100% native, 855 calls)
• Parse failure rate: 3.2%
• Extraction methods: bracket_fallback: 31, json_array_direct: 56, json_object_ads_key: 349, json_object_no_ads: 262, json_object_segments_key: 10, json_object_single_ad: 110, json_object_single_ad_truncated: 3, parse_failure: 27, regex_json_array: 7
• Verbosity: 9/855 calls over 1024 output tokens (1.1%); 0 hit max_tokens (0.0%); 3 salvaged from truncated JSON (0.4%)
• Segment category named on 473/479 detections (99%); the rest stay uncategorized (resolver: production)

xiaomi/mimo-v2.5-pro

• F1 (avg across episodes): 0.438
• Total cost / episode: $0.4954
• p50 / p95 latency: 2.19s / 6.41s
• JSON compliance: 0.90
• JSON mode: native (100% native, 855 calls)
• Parse failure rate: 2.6%
• Extraction methods: json_object_ads_detected_key: 1, json_object_ads_key: 133, json_object_advertisement_segments_key: 7, json_object_no_ads: 552, json_object_segments_key: 11, json_object_single_ad: 119, json_object_single_ad_truncated: 5, parse_failure: 22, regex_json_array: 5
• Verbosity: 9/855 calls over 1024 output tokens (1.1%); 1 hit max_tokens (0.1%); 5 salvaged from truncated JSON (0.6%)
• Segment category named on 267/274 detections (97%); the rest stay uncategorized (resolver: production)

cohere/command-a

• F1 (avg across episodes): 0.425
• Total cost / episode: $2.9415
• p50 / p95 latency: 1.72s / 2.92s
• JSON compliance: 0.71
• JSON mode: native (100% native, 855 calls)
• Parse failure rate: 0.0%
• Extraction methods: json_object_no_ads: 21, json_object_single_ad: 834
• Verbosity: 0/855 calls over 1024 output tokens (0.0%); 0 hit max_tokens (0.0%); 0 salvaged from truncated JSON (0.0%)
• Segment category named on 653/655 detections (100%); the rest stay uncategorized (resolver: production)

moonshotai/kimi-k2.6

• F1 (avg across episodes): 0.401
• Total cost / episode: $1.8854
• p50 / p95 latency: 55.69s / 164.28s
• JSON compliance: 0.63
• JSON mode: native (100% native, 855 calls)
• Parse failure rate: 29.7%
• Extraction methods: json_array_direct: 270, json_object_ads_key: 57, json_object_no_ads: 77, json_object_segments_key: 3, json_object_single_ad: 188, json_object_single_ad_truncated: 6, parse_failure: 254
• Verbosity: 826/855 calls over 1024 output tokens (96.6%); 260 hit max_tokens (30.4%); 6 salvaged from truncated JSON (0.7%)
• Segment category named on 194/202 detections (96%); the rest stay uncategorized (resolver: production)

stepfun/step-3.7-flash

• F1 (avg across episodes): 0.398
• Total cost / episode: $0.4302
• p50 / p95 latency: 13.87s / 35.00s
• JSON compliance: 0.56
• JSON mode: native (100% native, 725 calls)
• Parse failure rate: 32.3%
• Extraction methods: json_object_ads_key: 380, json_object_no_ads: 4, json_object_single_ad: 103, json_object_single_ad_truncated: 4, parse_failure: 234
• Verbosity: 388/725 calls over 1024 output tokens (53.5%); 238 hit max_tokens (32.8%); 4 salvaged from truncated JSON (0.6%)
• Segment category named on 89/92 detections (97%); the rest stay uncategorized (resolver: production)

meituan/longcat-2.0

• F1 (avg across episodes): 0.390
• Total cost / episode: $0.5573
• p50 / p95 latency: 10.53s / 77.08s
• JSON compliance: 0.42
• JSON mode: native (100% native, 855 calls)
• Parse failure rate: 50.2%
• Extraction methods: json_object_ads_key: 414, json_object_no_ads: 8, json_object_single_ad: 1, json_object_single_ad_truncated: 2, parse_failure: 429, regex_json_array: 1
• Verbosity: 287/855 calls over 1024 output tokens (33.6%); 92 hit max_tokens (10.8%); 2 salvaged from truncated JSON (0.2%)
• Segment category named on 263/265 detections (99%); the rest stay uncategorized (resolver: production)

deepseek/deepseek-r1-distill-llama-70b

• F1 (avg across episodes): 0.387
• Total cost / episode: $0.6913
• p50 / p95 latency: 27.21s / 97.56s
• JSON compliance: 0.39
• JSON mode: prompt-inject (0% native, 855 calls)
• Parse failure rate: 15.8%
• Extraction methods: json_array_direct: 36, json_object_single_ad_truncated: 1, markdown_code_block: 110, parse_failure: 135, regex_json_array: 573
• Verbosity: 241/855 calls over 1024 output tokens (28.2%); 57 hit max_tokens (6.7%); 1 salvaged from truncated JSON (0.1%)
• Segment category named on 360/385 detections (94%); the rest stay uncategorized (resolver: production)

qwen/qwen3-14b

• F1 (avg across episodes): 0.385
• Total cost / episode: $0.3989
• p50 / p95 latency: 14.39s / 59.03s
• JSON compliance: 0.60
• JSON mode: native (100% native, 855 calls)
• Parse failure rate: 15.3%
• Extraction methods: json_object_no_ads: 10, json_object_single_ad: 712, json_object_single_ad_truncated: 2, parse_failure: 131
• Verbosity: 155/855 calls over 1024 output tokens (18.1%); 23 hit max_tokens (2.7%); 2 salvaged from truncated JSON (0.2%)
• Segment category named on 318/318 detections (100%); the rest stay uncategorized (resolver: production)

cohere/command-r-plus-08-2024

• F1 (avg across episodes): 0.381
• Total cost / episode: $2.8505
• p50 / p95 latency: 1.13s / 16.24s
• JSON compliance: 0.93
• JSON mode: native (100% native, 855 calls)
• Parse failure rate: 0.0%
• Extraction methods: json_object_ads_key: 5, json_object_no_ads: 661, json_object_single_ad: 189
• Verbosity: 0/855 calls over 1024 output tokens (0.0%); 0 hit max_tokens (0.0%); 0 salvaged from truncated JSON (0.0%)
• Segment category named on 196/196 detections (100%); the rest stay uncategorized (resolver: production)

thinkingmachines/inkling-small

• F1 (avg across episodes): 0.358
• Total cost / episode: $1.0333
• p50 / p95 latency: 18.51s / 30.60s
• JSON compliance: 0.58
• JSON mode: native (100% native, 855 calls)
• Parse failure rate: 29.2%
• Extraction methods: json_object_no_ads: 281, json_object_single_ad: 284, json_object_single_ad_truncated: 4, parse_failure: 250, regex_json_array: 36
• Verbosity: 715/855 calls over 1024 output tokens (83.6%); 290 hit max_tokens (33.9%); 4 salvaged from truncated JSON (0.5%)
• Segment category named on 251/260 detections (97%); the rest stay uncategorized (resolver: production)

inclusionai/ring-2.6-1t

• F1 (avg across episodes): 0.337
• Total cost / episode: $0.2468
• p50 / p95 latency: 10.75s / 35.76s
• JSON compliance: 0.81
• JSON mode: native (100% native, 855 calls)
• Parse failure rate: 12.9%
• Extraction methods: bracket_fallback: 1, json_object_no_ads: 568, json_object_segments_key: 6, json_object_single_ad: 168, json_object_single_ad_truncated: 2, parse_failure: 110
• Verbosity: 479/855 calls over 1024 output tokens (56.0%); 111 hit max_tokens (13.0%); 2 salvaged from truncated JSON (0.2%)
• Segment category named on 167/169 detections (99%); the rest stay uncategorized (resolver: production)

openai/gpt-3.5-turbo

• F1 (avg across episodes): 0.310
• Total cost / episode: $0.5564
• p50 / p95 latency: 1.38s / 1.96s
• JSON compliance: 0.70
• JSON mode: native (100% native, 855 calls)
• Parse failure rate: 0.6%
• Extraction methods: json_object_no_ads: 19, json_object_single_ad: 831, parse_failure: 5
• Verbosity: 0/855 calls over 1024 output tokens (0.0%); 0 hit max_tokens (0.0%); 0 salvaged from truncated JSON (0.0%)
• Segment category named on 767/767 detections (100%); the rest stay uncategorized (resolver: production)

microsoft/phi-4

• F1 (avg across episodes): 0.299
• Total cost / episode: $0.0792
• p50 / p95 latency: 0.49s / 5.50s
• JSON compliance: 0.98
• JSON mode: native (100% native, 855 calls)
• Parse failure rate: 1.5%
• Extraction methods: json_array_direct: 828, json_object_no_ads: 4, json_object_single_ad: 3, json_object_single_ad_truncated: 7, parse_failure: 13
• Verbosity: 21/855 calls over 1024 output tokens (2.5%); 20 hit max_tokens (2.3%); 7 salvaged from truncated JSON (0.8%)
• Segment category named on 292/298 detections (98%); the rest stay uncategorized (resolver: production)

qwen/qwen3.8-max

• F1 (avg across episodes): 0.272
• Total cost / episode: $4.2714
• p50 / p95 latency: 29.66s / 90.22s
• JSON compliance: 0.73
• JSON mode: native (100% native, 855 calls)
• Parse failure rate: 26.9%
• Extraction methods: json_array_direct: 621, json_object_single_ad: 3, json_object_single_ad_truncated: 1, parse_failure: 230
• Verbosity: 456/855 calls over 1024 output tokens (53.3%); 231 hit max_tokens (27.0%); 1 salvaged from truncated JSON (0.1%)
• Segment category named on 113/114 detections (99%); the rest stay uncategorized (resolver: production)

meta-llama/llama-3.1-8b-instruct

• F1 (avg across episodes): 0.259
• Total cost / episode: $0.0572
• p50 / p95 latency: 0.93s / 4.34s
• JSON compliance: 0.82
• JSON mode: native (100% native, 855 calls)
• Parse failure rate: 0.0%
• Extraction methods: json_array_direct: 208, json_object_no_ads: 138, json_object_single_ad: 509
• Verbosity: 45/855 calls over 1024 output tokens (5.3%); 5 hit max_tokens (0.6%); 0 salvaged from truncated JSON (0.0%)
• Segment category named on 1525/1525 detections (100%); the rest stay uncategorized (resolver: production)

thinkingmachines/inkling

• F1 (avg across episodes): 0.225
• Total cost / episode: $3.2258
• p50 / p95 latency: 65.91s / 324.88s
• JSON compliance: 0.35
• JSON mode: native (100% native, 846 calls)
• Parse failure rate: 50.8%
• Extraction methods: bracket_fallback: 81, json_object_ads_key: 3, json_object_no_ads: 142, json_object_single_ad: 188, json_object_single_ad_truncated: 2, parse_failure: 430
• Verbosity: 703/846 calls over 1024 output tokens (83.1%); 505 hit max_tokens (59.7%); 2 salvaged from truncated JSON (0.2%)
• Segment category named on 77/79 detections (97%); the rest stay uncategorized (resolver: production)

openai/o4-mini

• F1 (avg across episodes): 0.135
• Total cost / episode: $2.1179
• p50 / p95 latency: 7.34s / 25.55s
• JSON compliance: 0.07
• JSON mode: native (100% native, 855 calls)
• Parse failure rate: 91.0%
• Extraction methods: json_object_no_ads: 24, json_object_segments_key: 1, json_object_single_ad: 52, parse_failure: 778
• Verbosity: 380/855 calls over 1024 output tokens (44.4%); 17 hit max_tokens (2.0%); 0 salvaged from truncated JSON (0.0%)
• Segment category named on 52/52 detections (100%); the rest stay uncategorized (resolver: production)

mistralai/mistral-small-2603

• F1 (avg across episodes): 0.019
• Total cost / episode: $0.1745
• p50 / p95 latency: 0.66s / 1.16s
• JSON compliance: 1.00
• JSON mode: native (100% native, 855 calls)
• Parse failure rate: 0.0%
• Extraction methods: json_array_direct: 855
• Verbosity: 0/855 calls over 1024 output tokens (0.0%); 0 hit max_tokens (0.0%); 0 salvaged from truncated JSON (0.0%)
• Segment category named on 20/20 detections (100%); the rest stay uncategorized (resolver: production)

qwen/qwen3-8b

• F1 (avg across episodes): 0.000
• Total cost / episode: $0.3256
• p50 / p95 latency: 39.80s / 170.51s
• JSON compliance: 0.10
• JSON mode: native (100% native, 855 calls)
• Parse failure rate: 88.9%
• Extraction methods: bracket_fallback: 15, json_array_direct: 1, json_object_no_ads: 79, parse_failure: 760
• Verbosity: 547/855 calls over 1024 output tokens (64.0%); 263 hit max_tokens (30.8%); 0 salvaged from truncated JSON (0.0%)
• Segment category named on 1/1 detections (100%); the rest stay uncategorized (resolver: production)

Per-Episode Detail

One subsection per episode in the corpus, showing how every model performed on that specific episode. For ad-bearing episodes you see F1 and the stdev across trials (low stdev means stable, high stdev means the model’s number on this episode is noisy). For the no-ad episode you see PASS / FAIL on the negative control: PASS = zero false positives across all windows, FAIL = the model flagged something that wasn’t an ad, with the count.

ep-ai-cloud-essentials-e8dc897fbd6b: How Physical AI is Streamlining Engineering

• Podcast: ai-cloud-essentials
• Duration: 16.4 min
• Truth: no-ads episode

Model	Result	FP count
claude-fable-5	PASS	0
claude-haiku-4-5-20251001	PASS	0
claude-opus-4-7	PASS	0
claude-opus-4-8	PASS	0
claude-opus-5	PASS	0
claude-sonnet-4-6	PASS	0
claude-sonnet-5	PASS	0
cohere/command-r-plus-08-2024	PASS	0
deepseek/deepseek-v3.2	PASS	0
deepseek/deepseek-v4-flash	PASS	0
deepseek/deepseek-v4-flash-0731	PASS	0
google/gemini-2.5-flash	PASS	0
google/gemini-3.1-pro-preview	PASS	0
google/gemini-3.5-flash	PASS	0
google/gemini-3.5-flash-lite	PASS	0
google/gemini-3.6-flash	PASS	0
inclusionai/ring-2.6-1t	PASS	0
meituan/longcat-2.0	PASS	0
meta-llama/llama-3.1-8b-instruct	PASS	0
meta-llama/llama-3.3-70b-instruct	PASS	0
meta/muse-spark-1.1	PASS	0
microsoft/phi-4	PASS	0
mistralai/codestral-2508	PASS	0
mistralai/mistral-medium-3-5	PASS	0
mistralai/mistral-medium-3.1	PASS	0
mistralai/mistral-small-2603	PASS	0
openai/gpt-5.6-luna	PASS	0
openai/o3	PASS	0
openai/o4-mini	PASS	0
qwen/qwen3-8b	PASS	0
qwen/qwen3.5-27b	PASS	0
qwen/qwen3.5-plus-02-15	PASS	0
qwen/qwen3.6-flash	PASS	0
qwen/qwen3.6-plus	PASS	0
qwen/qwen3.7-flash	PASS	0
qwen/qwen3.7-max	PASS	0
qwen/qwen3.8-max	PASS	0
stepfun/step-3.7-flash	PASS	0
tencent/hy3	PASS	0
thinkingmachines/inkling	PASS	0
x-ai/grok-4.3	PASS	0
x-ai/grok-4.5	PASS	0
xiaomi/mimo-v2.5-pro	PASS	0
deepseek/deepseek-r1	FAIL	1
deepseek/deepseek-r1-0528	FAIL	1
deepseek/deepseek-r1-distill-llama-70b	FAIL	1
deepseek/deepseek-v4-pro	FAIL	1
google/gemini-2.5-pro	FAIL	1
google/gemini-3.1-flash-lite	FAIL	1
google/gemma-4-26b-a4b-it	FAIL	1
google/gemma-4-31b-it	FAIL	1
meta-llama/llama-4-maverick	FAIL	1
meta-llama/llama-4-scout	FAIL	1
minimax/minimax-m3	FAIL	1
mistralai/mistral-large-2512	FAIL	1
moonshotai/kimi-k2.6	FAIL	1
moonshotai/kimi-k3	FAIL	1
nvidia/nemotron-3-super-120b-a12b	FAIL	1
nvidia/nemotron-3-ultra-550b-a55b	FAIL	1
openai/gpt-5.4	FAIL	1
openai/gpt-5.5	FAIL	1
openai/gpt-oss-120b	FAIL	1
openai/gpt-oss-20b	FAIL	1
qwen/qwen3-14b	FAIL	1
qwen/qwen3-235b-a22b-2507	FAIL	1
thinkingmachines/inkling-small	FAIL	1
google/gemini-2.5-flash-lite	FAIL	2
openai/gpt-3.5-turbo	FAIL	2
openai/gpt-5.4-mini	FAIL	2
openai/gpt-5.6-terra	FAIL	2
qwen/qwen3.7-plus	FAIL	2
xiaomi/mimo-v2.5	FAIL	2
z-ai/glm-5.2	FAIL	2
cohere/command-a	FAIL	3
openai/gpt-5.6-sol	FAIL	3

ep-crime-junkie-8ce498f299d7: MISSING: Christopher “Cole” Thomas

• Podcast: crime-junkie
• Duration: 48.2 min
• Truth ads: 4

Model	F1	F1 stdev
claude-haiku-4-5-20251001	1.000	0.000
claude-opus-5	1.000	0.000
claude-sonnet-4-6	1.000	0.000
deepseek/deepseek-v4-flash	1.000	0.000
google/gemini-2.5-flash	1.000	0.000
google/gemini-3.1-pro-preview	1.000	0.000
google/gemini-3.5-flash	1.000	0.000
google/gemini-3.6-flash	1.000	0.000
mistralai/mistral-medium-3-5	1.000	0.000
openai/gpt-5.5	1.000	0.000
openai/gpt-5.6-sol	1.000	0.000
openai/gpt-5.6-terra	1.000	0.000
qwen/qwen3.5-plus-02-15	1.000	0.000
qwen/qwen3.6-flash	1.000	0.000
qwen/qwen3.7-flash	1.000	0.000
x-ai/grok-4.3	1.000	0.000
claude-opus-4-7	0.978	0.050
mistralai/mistral-medium-3.1	0.978	0.050
nvidia/nemotron-3-super-120b-a12b	0.971	0.064
qwen/qwen3.6-plus	0.971	0.064
openai/gpt-5.4-mini	0.956	0.061
x-ai/grok-4.5	0.956	0.061
google/gemini-3.1-flash-lite	0.950	0.112
google/gemma-4-26b-a4b-it	0.950	0.112
qwen/qwen3.7-max	0.950	0.112
google/gemini-2.5-flash-lite	0.943	0.078
openai/gpt-oss-120b	0.943	0.078
openai/gpt-oss-20b	0.943	0.078
claude-sonnet-5	0.938	0.091
claude-fable-5	0.933	0.061
google/gemini-2.5-pro	0.933	0.061
moonshotai/kimi-k3	0.933	0.061
openai/gpt-5.6-luna	0.933	0.061
mistralai/mistral-large-2512	0.933	0.149
z-ai/glm-5.2	0.910	0.124
deepseek/deepseek-v4-flash-0731	0.905	0.147
claude-opus-4-8	0.889	0.000
meta-llama/llama-4-maverick	0.889	0.000
stepfun/step-3.7-flash	0.886	0.064
google/gemini-3.5-flash-lite	0.876	0.137
google/gemma-4-31b-it	0.876	0.137
deepseek/deepseek-r1	0.857	0.175
deepseek/deepseek-v3.2	0.857	0.000
deepseek/deepseek-v4-pro	0.851	0.260
xiaomi/mimo-v2.5	0.843	0.151
openai/gpt-5.4	0.818	0.040
nvidia/nemotron-3-ultra-550b-a55b	0.813	0.162
meta/muse-spark-1.1	0.813	0.272
openai/gpt-3.5-turbo	0.803	0.057
cohere/command-a	0.800	0.000
meta-llama/llama-3.3-70b-instruct	0.800	0.209
mistralai/codestral-2508	0.800	0.112
tencent/hy3	0.781	0.104
qwen/qwen3.7-plus	0.744	0.091
qwen/qwen3-235b-a22b-2507	0.731	0.152
deepseek/deepseek-r1-0528	0.728	0.097
qwen/qwen3-14b	0.690	0.163
xiaomi/mimo-v2.5-pro	0.651	0.163
cohere/command-r-plus-08-2024	0.613	0.119
meta-llama/llama-4-scout	0.600	0.091
moonshotai/kimi-k2.6	0.581	0.144
thinkingmachines/inkling-small	0.530	0.314
inclusionai/ring-2.6-1t	0.507	0.146
minimax/minimax-m3	0.507	0.146
openai/o3	0.507	0.146
deepseek/deepseek-r1-distill-llama-70b	0.373	0.037
qwen/qwen3.5-27b	0.373	0.239
thinkingmachines/inkling	0.307	0.174
qwen/qwen3.8-max	0.280	0.284
microsoft/phi-4	0.237	0.229
meta-llama/llama-3.1-8b-instruct	0.200	0.209
openai/o4-mini	0.160	0.219
meituan/longcat-2.0	0.000	0.000
mistralai/mistral-small-2603	0.000	0.000
qwen/qwen3-8b	0.000	0.000

ep-daily-gist-chicago-70a82fe93a5c: Suburban apartment market heats up

• Podcast: daily-gist-chicago
• Duration: 21.2 min
• Truth ads: 2

Model	F1	F1 stdev
claude-haiku-4-5-20251001	1.000	0.000
google/gemini-3.5-flash-lite	1.000	0.000
google/gemma-4-26b-a4b-it	0.920	0.110
inclusionai/ring-2.6-1t	0.867	0.183
openai/gpt-oss-120b	0.820	0.205
claude-opus-4-8	0.800	0.000
google/gemini-2.5-flash-lite	0.800	0.000
google/gemini-3.1-flash-lite	0.800	0.000
mistralai/mistral-medium-3-5	0.800	0.000
openai/gpt-5.4-mini	0.800	0.000
openai/gpt-oss-20b	0.780	0.179
nvidia/nemotron-3-super-120b-a12b	0.760	0.219
mistralai/mistral-medium-3.1	0.747	0.073
stepfun/step-3.7-flash	0.733	0.435
claude-fable-5	0.700	0.245
qwen/qwen3-14b	0.680	0.164
deepseek/deepseek-v4-flash	0.673	0.239
deepseek/deepseek-v4-flash-0731	0.667	0.408
qwen/qwen3.5-27b	0.640	0.434
deepseek/deepseek-v3.2	0.633	0.075
claude-sonnet-4-6	0.600	0.224
deepseek/deepseek-r1-distill-llama-70b	0.593	0.379
meta-llama/llama-3.1-8b-instruct	0.593	0.379
meta-llama/llama-3.3-70b-instruct	0.567	0.091
claude-sonnet-5	0.560	0.219
minimax/minimax-m3	0.533	0.298
google/gemini-3.1-pro-preview	0.500	0.000
google/gemini-3.5-flash	0.500	0.000
google/gemini-3.6-flash	0.500	0.000
qwen/qwen3.5-plus-02-15	0.500	0.000
qwen/qwen3.6-flash	0.500	0.000
qwen/qwen3.7-flash	0.500	0.000
qwen/qwen3.7-max	0.500	0.000
x-ai/grok-4.5	0.500	0.000
qwen/qwen3-235b-a22b-2507	0.480	0.179
moonshotai/kimi-k3	0.480	0.045
google/gemini-2.5-flash	0.473	0.116
nvidia/nemotron-3-ultra-550b-a55b	0.473	0.306
x-ai/grok-4.3	0.460	0.055
deepseek/deepseek-r1-0528	0.440	0.358
meta/muse-spark-1.1	0.433	0.435
qwen/qwen3.6-plus	0.433	0.253
openai/gpt-5.4	0.420	0.045
openai/gpt-5.6-sol	0.420	0.045
meta-llama/llama-4-scout	0.413	0.433
mistralai/mistral-large-2512	0.407	0.060
claude-opus-4-7	0.400	0.000
claude-opus-5	0.400	0.000
cohere/command-a	0.400	0.000
google/gemini-2.5-pro	0.400	0.000
google/gemma-4-31b-it	0.400	0.000
mistralai/codestral-2508	0.400	0.224
openai/gpt-3.5-turbo	0.400	0.000
openai/gpt-5.5	0.400	0.000
openai/gpt-5.6-luna	0.400	0.000
openai/gpt-5.6-terra	0.400	0.000
qwen/qwen3.7-plus	0.400	0.000
xiaomi/mimo-v2.5-pro	0.400	0.365
deepseek/deepseek-r1	0.360	0.207
meta-llama/llama-4-maverick	0.320	0.179
z-ai/glm-5.2	0.320	0.179
deepseek/deepseek-v4-pro	0.280	0.259
cohere/command-r-plus-08-2024	0.267	0.365
meituan/longcat-2.0	0.233	0.325
thinkingmachines/inkling-small	0.160	0.219
openai/o4-mini	0.133	0.298
tencent/hy3	0.100	0.224
moonshotai/kimi-k2.6	0.080	0.179
xiaomi/mimo-v2.5	0.080	0.179
microsoft/phi-4	0.000	0.000
mistralai/mistral-small-2603	0.000	0.000
openai/o3	0.000	0.000
qwen/qwen3-8b	0.000	0.000
qwen/qwen3.8-max	0.000	0.000
thinkingmachines/inkling	0.000	0.000

ep-daily-tech-news-show-b576979e1fe8: Motorola Razr Fold is a Noble Competitor to the Galaxy Z Fold 7 - DTNS 5269

• Podcast: daily-tech-news-show
• Duration: 34.6 min
• Truth ads: 4

Model	F1	F1 stdev
claude-haiku-4-5-20251001	1.000	0.000
claude-opus-4-7	1.000	0.000
claude-opus-5	1.000	0.000
claude-sonnet-4-6	1.000	0.000
google/gemini-3.5-flash	1.000	0.000
google/gemini-3.6-flash	1.000	0.000
meta-llama/llama-4-maverick	1.000	0.000
mistralai/mistral-medium-3-5	1.000	0.000
openai/gpt-5.4-mini	1.000	0.000
openai/gpt-5.5	1.000	0.000
openai/gpt-5.6-luna	1.000	0.000
openai/gpt-5.6-sol	1.000	0.000
qwen/qwen3.7-max	1.000	0.000
x-ai/grok-4.5	1.000	0.000
claude-fable-5	0.978	0.050
x-ai/grok-4.3	0.978	0.050
google/gemini-2.5-flash-lite	0.971	0.064
claude-opus-4-8	0.956	0.061
google/gemini-3.1-flash-lite	0.956	0.061
google/gemini-3.1-pro-preview	0.950	0.112
openai/gpt-5.6-terra	0.950	0.112
google/gemini-2.5-pro	0.943	0.078
moonshotai/kimi-k3	0.943	0.078
qwen/qwen3.6-flash	0.921	0.114
claude-sonnet-5	0.911	0.145
mistralai/mistral-large-2512	0.893	0.071
google/gemma-4-31b-it	0.893	0.107
google/gemini-3.5-flash-lite	0.892	0.062
google/gemini-2.5-flash	0.889	0.000
qwen/qwen3.5-plus-02-15	0.889	0.000
deepseek/deepseek-v4-flash	0.886	0.064
openai/gpt-5.4	0.878	0.125
deepseek/deepseek-v4-pro	0.876	0.137
openai/gpt-oss-120b	0.855	0.149
deepseek/deepseek-r1	0.848	0.119
moonshotai/kimi-k2.6	0.819	0.195
z-ai/glm-5.2	0.813	0.162
mistralai/codestral-2508	0.800	0.112
nvidia/nemotron-3-super-120b-a12b	0.785	0.255
meta-llama/llama-3.3-70b-instruct	0.774	0.126
qwen/qwen3.7-flash	0.771	0.192
meta/muse-spark-1.1	0.771	0.152
deepseek/deepseek-r1-0528	0.744	0.198
tencent/hy3	0.743	0.104
qwen/qwen3-235b-a22b-2507	0.709	0.206
deepseek/deepseek-v4-flash-0731	0.706	0.189
qwen/qwen3.6-plus	0.686	0.104
deepseek/deepseek-v3.2	0.667	0.000
cohere/command-a	0.667	0.000
qwen/qwen3.7-plus	0.667	0.000
meta-llama/llama-4-scout	0.651	0.163
xiaomi/mimo-v2.5	0.649	0.175
meituan/longcat-2.0	0.613	0.119
qwen/qwen3.5-27b	0.598	0.197
thinkingmachines/inkling-small	0.592	0.206
minimax/minimax-m3	0.547	0.166
nvidia/nemotron-3-ultra-550b-a55b	0.547	0.117
openai/o3	0.545	0.209
qwen/qwen3-14b	0.529	0.039
inclusionai/ring-2.6-1t	0.522	0.118
xiaomi/mimo-v2.5-pro	0.518	0.332
mistralai/mistral-medium-3.1	0.444	0.000
google/gemma-4-26b-a4b-it	0.428	0.103
microsoft/phi-4	0.427	0.138
thinkingmachines/inkling	0.427	0.273
openai/gpt-oss-20b	0.421	0.144
qwen/qwen3.8-max	0.320	0.179
meta-llama/llama-3.1-8b-instruct	0.274	0.181
stepfun/step-3.7-flash	0.240	0.219
openai/gpt-3.5-turbo	0.222	0.000
deepseek/deepseek-r1-distill-llama-70b	0.160	0.219
openai/o4-mini	0.147	0.202
cohere/command-r-plus-08-2024	0.133	0.183
mistralai/mistral-small-2603	0.000	0.000
qwen/qwen3-8b	0.000	0.000

ep-daily-tech-news-show-c1904b8605f7: Switch 2 Prices Rise, Forecast Drops - DTNS 5265

• Podcast: daily-tech-news-show
• Duration: 38.6 min
• Truth ads: 5

Model	F1	F1 stdev
claude-opus-4-7	0.767	0.130
openai/gpt-5.5	0.667	0.000
openai/gpt-5.4	0.653	0.030
openai/gpt-5.6-terra	0.653	0.030
openai/gpt-3.5-turbo	0.622	0.099
deepseek/deepseek-v3.2	0.616	0.052
openai/gpt-5.6-luna	0.609	0.096
qwen/qwen3.6-flash	0.602	0.043
qwen/qwen3.7-flash	0.602	0.043
openai/gpt-5.6-sol	0.600	0.115
google/gemini-3.6-flash	0.589	0.024
mistralai/mistral-medium-3-5	0.589	0.024
deepseek/deepseek-v4-pro	0.581	0.084
google/gemini-3.1-pro-preview	0.578	0.030
google/gemini-3.5-flash	0.578	0.030
moonshotai/kimi-k3	0.578	0.030
x-ai/grok-4.3	0.578	0.030
xiaomi/mimo-v2.5	0.569	0.135
claude-haiku-4-5-20251001	0.567	0.030
google/gemini-3.1-flash-lite	0.567	0.030
claude-opus-5	0.564	0.094
claude-sonnet-4-6	0.556	0.024
google/gemini-2.5-flash-lite	0.545	0.000
mistralai/mistral-large-2512	0.545	0.000
qwen/qwen3.5-plus-02-15	0.545	0.000
openai/gpt-5.4-mini	0.533	0.122
claude-fable-5	0.529	0.044
claude-opus-4-8	0.527	0.025
nvidia/nemotron-3-ultra-550b-a55b	0.512	0.158
x-ai/grok-4.5	0.509	0.020
z-ai/glm-5.2	0.508	0.147
google/gemini-3.5-flash-lite	0.502	0.113
google/gemini-2.5-flash	0.500	0.000
claude-sonnet-5	0.492	0.017
google/gemini-2.5-pro	0.490	0.039
meta-llama/llama-4-maverick	0.480	0.110
google/gemma-4-31b-it	0.476	0.070
openai/gpt-oss-120b	0.475	0.136
qwen/qwen3.7-plus	0.468	0.144
qwen/qwen3.7-max	0.462	0.000
meta/muse-spark-1.1	0.456	0.072
meituan/longcat-2.0	0.436	0.158
cohere/command-a	0.436	0.020
qwen/qwen3-235b-a22b-2507	0.436	0.020
mistralai/mistral-medium-3.1	0.429	0.133
minimax/minimax-m3	0.429	0.044
deepseek/deepseek-r1-distill-llama-70b	0.422	0.115
nvidia/nemotron-3-super-120b-a12b	0.417	0.181
qwen/qwen3.5-27b	0.417	0.096
xiaomi/mimo-v2.5-pro	0.414	0.138
deepseek/deepseek-r1-0528	0.411	0.098
deepseek/deepseek-v4-flash	0.402	0.089
cohere/command-r-plus-08-2024	0.361	0.115
openai/o3	0.356	0.210
deepseek/deepseek-r1	0.352	0.108
qwen/qwen3.6-plus	0.336	0.242
mistralai/codestral-2508	0.307	0.126
deepseek/deepseek-v4-flash-0731	0.295	0.021
meta-llama/llama-3.3-70b-instruct	0.295	0.104
openai/gpt-oss-20b	0.279	0.157
meta-llama/llama-4-scout	0.256	0.046
microsoft/phi-4	0.239	0.015
qwen/qwen3.8-max	0.231	0.132
tencent/hy3	0.229	0.128
moonshotai/kimi-k2.6	0.181	0.166
google/gemma-4-26b-a4b-it	0.170	0.184
thinkingmachines/inkling-small	0.150	0.224
meta-llama/llama-3.1-8b-instruct	0.133	0.199
thinkingmachines/inkling	0.124	0.170
qwen/qwen3-14b	0.089	0.122
openai/o4-mini	0.067	0.149
inclusionai/ring-2.6-1t	0.057	0.128
mistralai/mistral-small-2603	0.000	0.000
qwen/qwen3-8b	0.000	0.000
stepfun/step-3.7-flash	0.000	0.000

ep-drink-champs-30c9a2d49f13: Episode 501 w/ Warren Sapp

• Podcast: drink-champs
• Duration: 258.6 min
• Truth ads: 9

Model	F1	F1 stdev
qwen/qwen3.5-plus-02-15	0.967	0.049
google/gemini-3.5-flash-lite	0.959	0.042
claude-haiku-4-5-20251001	0.937	0.086
x-ai/grok-4.5	0.899	0.023
claude-sonnet-4-6	0.891	0.005
google/gemini-3.5-flash	0.873	0.055
claude-opus-4-8	0.870	0.026
claude-opus-5	0.863	0.041
deepseek/deepseek-v3.2	0.859	0.038
google/gemini-3.1-flash-lite	0.858	0.029
google/gemini-3.6-flash	0.842	0.000
google/gemini-3.1-pro-preview	0.834	0.019
mistralai/mistral-medium-3-5	0.833	0.064
google/gemini-2.5-pro	0.829	0.029
claude-opus-4-7	0.825	0.037
openai/gpt-5.5	0.823	0.055
mistralai/mistral-medium-3.1	0.821	0.047
x-ai/grok-4.3	0.814	0.168
claude-sonnet-5	0.806	0.070
google/gemini-2.5-flash	0.800	0.041
minimax/minimax-m3	0.798	0.105
openai/gpt-5.6-luna	0.778	0.000
openai/gpt-5.6-terra	0.774	0.064
claude-fable-5	0.770	0.017
openai/gpt-5.4	0.706	0.110
deepseek/deepseek-v4-flash	0.687	0.079
google/gemma-4-31b-it	0.668	0.131
google/gemini-2.5-flash-lite	0.657	0.160
openai/gpt-5.6-sol	0.648	0.089
qwen/qwen3.6-flash	0.641	0.085
qwen/qwen3.7-max	0.628	0.187
qwen/qwen3.5-27b	0.625	0.089
meta/muse-spark-1.1	0.604	0.230
mistralai/mistral-large-2512	0.574	0.021
openai/gpt-5.4-mini	0.563	0.123
qwen/qwen3.7-flash	0.546	0.224
z-ai/glm-5.2	0.543	0.069
deepseek/deepseek-v4-pro	0.532	0.190
qwen/qwen3.6-plus	0.470	0.331
openai/o3	0.464	0.202
mistralai/codestral-2508	0.452	0.091
deepseek/deepseek-r1-0528	0.422	0.235
moonshotai/kimi-k3	0.422	0.094
google/gemma-4-26b-a4b-it	0.412	0.213
meituan/longcat-2.0	0.400	0.163
qwen/qwen3.7-plus	0.394	0.046
xiaomi/mimo-v2.5	0.384	0.096
xiaomi/mimo-v2.5-pro	0.383	0.152
deepseek/deepseek-v4-flash-0731	0.365	0.195
deepseek/deepseek-r1	0.365	0.082
deepseek/deepseek-r1-distill-llama-70b	0.312	0.059
openai/gpt-oss-120b	0.305	0.100
meta-llama/llama-4-maverick	0.289	0.006
qwen/qwen3-235b-a22b-2507	0.186	0.152
meta-llama/llama-3.1-8b-instruct	0.167	0.041
openai/gpt-oss-20b	0.147	0.087
cohere/command-r-plus-08-2024	0.139	0.078
thinkingmachines/inkling-small	0.133	0.122
nvidia/nemotron-3-ultra-550b-a55b	0.091	0.124
meta-llama/llama-4-scout	0.085	0.062
openai/gpt-3.5-turbo	0.055	0.001
thinkingmachines/inkling	0.042	0.083
meta-llama/llama-3.3-70b-instruct	0.040	0.089
moonshotai/kimi-k2.6	0.036	0.081
tencent/hy3	0.033	0.075
inclusionai/ring-2.6-1t	0.031	0.069
nvidia/nemotron-3-super-120b-a12b	0.029	0.064
qwen/qwen3-14b	0.018	0.041
cohere/command-a	0.000	0.000
microsoft/phi-4	0.000	0.000
mistralai/mistral-small-2603	0.000	0.000
openai/o4-mini	0.000	0.000
qwen/qwen3-8b	0.000	0.000
qwen/qwen3.8-max	0.000	0.000

ep-glt1412515089-373d5ba5007b: #2496 - Julia Mossbridge

• Podcast: glt1412515089
• Duration: 165.3 min
• Truth ads: 4

Model	F1	F1 stdev
google/gemini-2.5-flash	1.000	0.000
deepseek/deepseek-r1-distill-llama-70b	0.971	0.064
openai/gpt-oss-120b	0.956	0.061
openai/gpt-5.6-luna	0.949	0.070
deepseek/deepseek-v4-flash	0.921	0.114
x-ai/grok-4.3	0.921	0.114
openai/gpt-5.5	0.914	0.078
nvidia/nemotron-3-super-120b-a12b	0.892	0.062
moonshotai/kimi-k3	0.886	0.064
google/gemma-4-26b-a4b-it	0.867	0.122
claude-fable-5	0.857	0.000
claude-opus-4-7	0.857	0.000
claude-opus-5	0.857	0.000
claude-sonnet-4-6	0.857	0.000
claude-sonnet-5	0.857	0.000
deepseek/deepseek-v4-pro	0.857	0.000
google/gemini-3.1-pro-preview	0.857	0.000
google/gemini-3.5-flash	0.857	0.000
google/gemini-3.6-flash	0.857	0.000
openai/gpt-5.6-terra	0.857	0.000
openai/o3	0.857	0.000
qwen/qwen3.5-plus-02-15	0.857	0.000
qwen/qwen3.7-max	0.857	0.000
x-ai/grok-4.5	0.857	0.000
minimax/minimax-m3	0.853	0.094
tencent/hy3	0.848	0.119
qwen/qwen3.6-plus	0.844	0.099
google/gemma-4-31b-it	0.828	0.114
claude-haiku-4-5-20251001	0.814	0.185
qwen/qwen3.5-27b	0.800	0.183
meituan/longcat-2.0	0.781	0.104
stepfun/step-3.7-flash	0.781	0.104
meta/muse-spark-1.1	0.766	0.204
qwen/qwen3.6-flash	0.755	0.068
google/gemini-2.5-pro	0.750	0.000
deepseek/deepseek-r1-0528	0.750	0.095
deepseek/deepseek-r1	0.746	0.152
claude-opus-4-8	0.743	0.156
deepseek/deepseek-v4-flash-0731	0.743	0.104
mistralai/mistral-medium-3-5	0.740	0.108
openai/gpt-5.6-sol	0.739	0.048
cohere/command-r-plus-08-2024	0.712	0.269
mistralai/mistral-medium-3.1	0.711	0.099
xiaomi/mimo-v2.5-pro	0.705	0.144
openai/gpt-oss-20b	0.692	0.075
google/gemini-3.1-flash-lite	0.681	0.047
qwen/qwen3.7-flash	0.664	0.374
meta-llama/llama-3.3-70b-instruct	0.659	0.073
deepseek/deepseek-v3.2	0.613	0.119
nvidia/nemotron-3-ultra-550b-a55b	0.598	0.208
openai/gpt-5.4	0.565	0.129
z-ai/glm-5.2	0.555	0.131
inclusionai/ring-2.6-1t	0.537	0.146
google/gemini-3.5-flash-lite	0.531	0.223
meta-llama/llama-4-maverick	0.518	0.025
qwen/qwen3-14b	0.488	0.148
moonshotai/kimi-k2.6	0.473	0.169
thinkingmachines/inkling-small	0.467	0.144
google/gemini-2.5-flash-lite	0.460	0.014
qwen/qwen3.7-plus	0.455	0.022
openai/gpt-5.4-mini	0.442	0.078
xiaomi/mimo-v2.5	0.402	0.031
mistralai/mistral-large-2512	0.400	0.080
qwen/qwen3.8-max	0.373	0.239
qwen/qwen3-235b-a22b-2507	0.372	0.218
mistralai/codestral-2508	0.371	0.016
meta-llama/llama-4-scout	0.367	0.342
cohere/command-a	0.345	0.011
openai/gpt-3.5-turbo	0.263	0.005
meta-llama/llama-3.1-8b-instruct	0.203	0.027
thinkingmachines/inkling	0.200	0.231
openai/o4-mini	0.147	0.202
microsoft/phi-4	0.000	0.000
mistralai/mistral-small-2603	0.000	0.000
qwen/qwen3-8b	0.000	0.000

ep-it-s-a-thing-e339179dfad6: SOUP shots - It’s a Thing 418

• Podcast: it-s-a-thing
• Duration: 26.7 min
• Truth ads: 1

Model	F1	F1 stdev
deepseek/deepseek-v4-pro	0.867	0.183
minimax/minimax-m3	0.800	0.447
google/gemini-2.5-pro	0.733	0.149
qwen/qwen3.6-flash	0.733	0.149
meta-llama/llama-3.1-8b-instruct	0.700	0.274
claude-fable-5	0.667	0.000
claude-haiku-4-5-20251001	0.667	0.000
claude-opus-4-7	0.667	0.000
claude-opus-4-8	0.667	0.000
claude-sonnet-4-6	0.667	0.000
google/gemini-3.1-flash-lite	0.667	0.000
google/gemini-3.5-flash	0.667	0.000
google/gemini-3.5-flash-lite	0.667	0.000
google/gemini-3.6-flash	0.667	0.000
mistralai/mistral-medium-3-5	0.667	0.000
openai/gpt-5.5	0.667	0.000
openai/gpt-5.6-luna	0.667	0.000
qwen/qwen3.5-plus-02-15	0.667	0.000
qwen/qwen3.7-max	0.667	0.000
x-ai/grok-4.3	0.667	0.000
x-ai/grok-4.5	0.667	0.000
claude-sonnet-5	0.633	0.075
google/gemini-3.1-pro-preview	0.633	0.075
openai/gpt-5.6-terra	0.600	0.091
qwen/qwen3.5-27b	0.600	0.365
qwen/qwen3.7-flash	0.600	0.365
z-ai/glm-5.2	0.547	0.117
google/gemma-4-31b-it	0.533	0.506
moonshotai/kimi-k3	0.533	0.298
claude-opus-5	0.500	0.000
mistralai/mistral-medium-3.1	0.500	0.000
deepseek/deepseek-v3.2	0.473	0.116
deepseek/deepseek-v4-flash	0.467	0.447
openai/gpt-5.6-sol	0.467	0.274
deepseek/deepseek-v4-flash-0731	0.400	0.548
meta-llama/llama-4-scout	0.400	0.548
microsoft/phi-4	0.400	0.365
mistralai/codestral-2508	0.400	0.000
openai/gpt-oss-120b	0.393	0.245
qwen/qwen3.7-plus	0.393	0.245
cohere/command-a	0.340	0.195
google/gemini-2.5-flash	0.333	0.000
meituan/longcat-2.0	0.333	0.471
nvidia/nemotron-3-ultra-550b-a55b	0.333	0.471
openai/gpt-5.4	0.333	0.312
openai/o3	0.333	0.471
mistralai/mistral-large-2512	0.320	0.057
google/gemini-2.5-flash-lite	0.309	0.051
openai/gpt-oss-20b	0.300	0.447
deepseek/deepseek-r1	0.267	0.365
meta-llama/llama-3.3-70b-instruct	0.267	0.365
meta/muse-spark-1.1	0.267	0.365
qwen/qwen3-235b-a22b-2507	0.233	0.325
xiaomi/mimo-v2.5	0.233	0.325
google/gemma-4-26b-a4b-it	0.133	0.298
deepseek/deepseek-r1-0528	0.120	0.179
openai/gpt-5.4-mini	0.100	0.224
xiaomi/mimo-v2.5-pro	0.100	0.224
cohere/command-r-plus-08-2024	0.000	0.000
deepseek/deepseek-r1-distill-llama-70b	0.000	0.000
inclusionai/ring-2.6-1t	0.000	0.000
meta-llama/llama-4-maverick	0.000	0.000
mistralai/mistral-small-2603	0.000	0.000
moonshotai/kimi-k2.6	0.000	0.000
nvidia/nemotron-3-super-120b-a12b	0.000	0.000
openai/gpt-3.5-turbo	0.000	0.000
openai/o4-mini	0.000	0.000
qwen/qwen3-14b	0.000	0.000
qwen/qwen3-8b	0.000	0.000
qwen/qwen3.6-plus	0.000	0.000
qwen/qwen3.8-max	0.000	0.000
stepfun/step-3.7-flash	0.000	0.000
tencent/hy3	0.000	0.000
thinkingmachines/inkling	0.000	0.000
thinkingmachines/inkling-small	0.000	0.000

ep-on-air-with-dan-and-alex2-574e4f303730: Ryanair Wants Alcohol Bans, Emirates’ $6.8B Record Profit & Buying Spirit Airlines?!

• Podcast: on-air-with-dan-and-alex2
• Duration: 58.1 min
• Truth ads: 2

Model	F1	F1 stdev
mistralai/mistral-medium-3-5	1.000	0.000
mistralai/mistral-medium-3.1	1.000	0.000
x-ai/grok-4.3	1.000	0.000
claude-fable-5	0.960	0.089
claude-sonnet-5	0.960	0.089
google/gemma-4-31b-it	0.960	0.089
openai/gpt-5.6-terra	0.960	0.089
qwen/qwen3.5-plus-02-15	0.960	0.089
x-ai/grok-4.5	0.960	0.089
google/gemini-3.5-flash-lite	0.933	0.149
claude-haiku-4-5-20251001	0.920	0.110
qwen/qwen3.8-max	0.900	0.224
qwen/qwen3.6-flash	0.840	0.089
qwen/qwen3.7-max	0.840	0.089
moonshotai/kimi-k2.6	0.813	0.119
claude-opus-4-8	0.800	0.000
claude-opus-5	0.800	0.000
deepseek/deepseek-r1	0.800	0.000
deepseek/deepseek-v4-flash	0.800	0.000
google/gemini-2.5-flash	0.800	0.000
google/gemini-2.5-pro	0.800	0.000
google/gemini-3.1-flash-lite	0.800	0.000
google/gemini-3.5-flash	0.800	0.000
microsoft/phi-4	0.800	0.183
mistralai/mistral-large-2512	0.800	0.000
nvidia/nemotron-3-super-120b-a12b	0.800	0.000
openai/gpt-5.4-mini	0.800	0.000
openai/gpt-5.5	0.800	0.000
qwen/qwen3.6-plus	0.800	0.000
qwen/qwen3.7-flash	0.800	0.000
minimax/minimax-m3	0.780	0.179
claude-sonnet-4-6	0.773	0.060
deepseek/deepseek-r1-0528	0.773	0.060
google/gemini-3.6-flash	0.773	0.060
tencent/hy3	0.753	0.185
openai/gpt-5.6-sol	0.747	0.073
claude-opus-4-7	0.740	0.134
deepseek/deepseek-v4-flash-0731	0.740	0.134
openai/o3	0.733	0.149
openai/gpt-oss-120b	0.728	0.105
google/gemini-2.5-flash-lite	0.720	0.179
qwen/qwen3-235b-a22b-2507	0.701	0.098
deepseek/deepseek-v4-pro	0.700	0.245
meta-llama/llama-4-maverick	0.693	0.060
openai/gpt-5.6-luna	0.693	0.060
qwen/qwen3.5-27b	0.693	0.213
thinkingmachines/inkling-small	0.693	0.060
openai/gpt-5.4	0.674	0.081
deepseek/deepseek-v3.2	0.667	0.000
google/gemini-3.1-pro-preview	0.667	0.000
google/gemma-4-26b-a4b-it	0.667	0.000
meta/muse-spark-1.1	0.667	0.408
moonshotai/kimi-k3	0.667	0.000
stepfun/step-3.7-flash	0.633	0.415
thinkingmachines/inkling	0.627	0.128
mistralai/codestral-2508	0.610	0.052
z-ai/glm-5.2	0.603	0.114
cohere/command-r-plus-08-2024	0.600	0.365
openai/gpt-oss-20b	0.560	0.146
qwen/qwen3.7-plus	0.514	0.032
nvidia/nemotron-3-ultra-550b-a55b	0.513	0.366
cohere/command-a	0.507	0.214
deepseek/deepseek-r1-distill-llama-70b	0.500	0.000
meituan/longcat-2.0	0.500	0.000
meta-llama/llama-4-scout	0.493	0.303
meta-llama/llama-3.3-70b-instruct	0.480	0.179
qwen/qwen3-14b	0.453	0.197
openai/gpt-3.5-turbo	0.444	0.000
xiaomi/mimo-v2.5	0.441	0.094
inclusionai/ring-2.6-1t	0.400	0.283
xiaomi/mimo-v2.5-pro	0.300	0.298
openai/o4-mini	0.267	0.365
meta-llama/llama-3.1-8b-instruct	0.263	0.283
mistralai/mistral-small-2603	0.000	0.000
qwen/qwen3-8b	0.000	0.000

ep-oxide-and-friends-ce789ff5b62e: Mechanical Engineering at Oxide [chapter images]

• Podcast: oxide-and-friends
• Duration: 84.5 min
• Truth: no-ads episode

Model	Result	FP count
claude-fable-5	PASS	0
claude-haiku-4-5-20251001	PASS	0
claude-opus-4-7	PASS	0
claude-opus-4-8	PASS	0
claude-opus-5	PASS	0
claude-sonnet-4-6	PASS	0
claude-sonnet-5	PASS	0
cohere/command-r-plus-08-2024	PASS	0
deepseek/deepseek-r1-distill-llama-70b	PASS	0
deepseek/deepseek-v3.2	PASS	0
deepseek/deepseek-v4-flash	PASS	0
deepseek/deepseek-v4-flash-0731	PASS	0
deepseek/deepseek-v4-pro	PASS	0
google/gemini-2.5-flash	PASS	0
google/gemini-2.5-flash-lite	PASS	0
google/gemini-3.1-flash-lite	PASS	0
google/gemini-3.1-pro-preview	PASS	0
google/gemini-3.5-flash	PASS	0
google/gemini-3.5-flash-lite	PASS	0
google/gemini-3.6-flash	PASS	0
google/gemma-4-26b-a4b-it	PASS	0
google/gemma-4-31b-it	PASS	0
inclusionai/ring-2.6-1t	PASS	0
meituan/longcat-2.0	PASS	0
meta-llama/llama-3.3-70b-instruct	PASS	0
meta-llama/llama-4-maverick	PASS	0
meta-llama/llama-4-scout	PASS	0
meta/muse-spark-1.1	PASS	0
microsoft/phi-4	PASS	0
mistralai/mistral-large-2512	PASS	0
mistralai/mistral-medium-3-5	PASS	0
mistralai/mistral-medium-3.1	PASS	0
mistralai/mistral-small-2603	PASS	0
moonshotai/kimi-k3	PASS	0
nvidia/nemotron-3-super-120b-a12b	PASS	0
openai/gpt-5.5	PASS	0
openai/gpt-5.6-luna	PASS	0
openai/o3	PASS	0
openai/o4-mini	PASS	0
qwen/qwen3-8b	PASS	0
qwen/qwen3.5-27b	PASS	0
qwen/qwen3.5-plus-02-15	PASS	0
qwen/qwen3.6-flash	PASS	0
qwen/qwen3.6-plus	PASS	0
qwen/qwen3.7-flash	PASS	0
qwen/qwen3.7-max	PASS	0
qwen/qwen3.8-max	PASS	0
stepfun/step-3.7-flash	PASS	0
tencent/hy3	PASS	0
x-ai/grok-4.3	PASS	0
x-ai/grok-4.5	PASS	0
google/gemini-2.5-pro	FAIL	1
minimax/minimax-m3	FAIL	1
mistralai/codestral-2508	FAIL	1
openai/gpt-5.6-terra	FAIL	1
openai/gpt-oss-120b	FAIL	1
openai/gpt-oss-20b	FAIL	1
qwen/qwen3-14b	FAIL	1
thinkingmachines/inkling	FAIL	1
xiaomi/mimo-v2.5-pro	FAIL	1
meta-llama/llama-3.1-8b-instruct	FAIL	2
openai/gpt-5.6-sol	FAIL	2
deepseek/deepseek-r1	FAIL	3
moonshotai/kimi-k2.6	FAIL	3
openai/gpt-5.4	FAIL	3
openai/gpt-5.4-mini	FAIL	3
nvidia/nemotron-3-ultra-550b-a55b	FAIL	4
cohere/command-a	FAIL	5
z-ai/glm-5.2	FAIL	5
qwen/qwen3-235b-a22b-2507	FAIL	6
thinkingmachines/inkling-small	FAIL	6
xiaomi/mimo-v2.5	FAIL	7
deepseek/deepseek-r1-0528	FAIL	8
openai/gpt-3.5-turbo	FAIL	11
qwen/qwen3.7-plus	FAIL	11

ep-security-now-audio-2850b24903b2: SN 1077: A Browser AI API? - End of Bug Bounties?

• Podcast: security-now-audio
• Duration: 156.2 min
• Truth ads: 6

Model	F1	F1 stdev
claude-opus-5	0.857	0.000
meta-llama/llama-3.3-70b-instruct	0.769	0.000
openai/gpt-5.5	0.769	0.000
deepseek/deepseek-v4-pro	0.764	0.084
openai/gpt-5.6-luna	0.758	0.025
qwen/qwen3.5-27b	0.751	0.153
cohere/command-r-plus-08-2024	0.751	0.033
qwen/qwen3-14b	0.744	0.099
nvidia/nemotron-3-super-120b-a12b	0.736	0.030
openai/gpt-5.6-sol	0.736	0.030
moonshotai/kimi-k3	0.728	0.056
openai/gpt-5.6-terra	0.716	0.036
claude-haiku-4-5-20251001	0.714	0.000
google/gemini-3.5-flash	0.714	0.000
google/gemma-4-26b-a4b-it	0.714	0.000
mistralai/mistral-medium-3-5	0.714	0.000
qwen/qwen3.6-flash	0.714	0.000
qwen/qwen3.7-flash	0.714	0.000
x-ai/grok-4.3	0.714	0.000
x-ai/grok-4.5	0.714	0.000
claude-opus-4-8	0.695	0.026
claude-sonnet-4-6	0.695	0.026
qwen/qwen3.7-max	0.695	0.026
google/gemini-3.5-flash-lite	0.695	0.044
claude-sonnet-5	0.693	0.060
claude-opus-4-7	0.693	0.038
qwen/qwen3.5-plus-02-15	0.686	0.064
openai/gpt-oss-20b	0.678	0.081
deepseek/deepseek-v4-flash	0.673	0.103
openai/gpt-oss-120b	0.668	0.032
claude-fable-5	0.667	0.000
google/gemini-3.1-pro-preview	0.667	0.000
google/gemini-3.6-flash	0.667	0.000
google/gemma-4-31b-it	0.667	0.000
tencent/hy3	0.664	0.064
google/gemini-2.5-pro	0.658	0.019
qwen/qwen3-235b-a22b-2507	0.652	0.095
openai/o3	0.635	0.169
google/gemini-2.5-flash	0.625	0.000
meta/muse-spark-1.1	0.623	0.026
deepseek/deepseek-v4-flash-0731	0.612	0.096
z-ai/glm-5.2	0.601	0.070
meta-llama/llama-4-maverick	0.596	0.016
qwen/qwen3.6-plus	0.592	0.052
openai/gpt-5.4	0.589	0.025
google/gemini-3.1-flash-lite	0.588	0.000
google/gemini-2.5-flash-lite	0.569	0.081
xiaomi/mimo-v2.5	0.563	0.138
minimax/minimax-m3	0.552	0.200
meta-llama/llama-4-scout	0.551	0.138
inclusionai/ring-2.6-1t	0.547	0.035
deepseek/deepseek-r1	0.543	0.078
deepseek/deepseek-r1-distill-llama-70b	0.542	0.048
moonshotai/kimi-k2.6	0.533	0.212
deepseek/deepseek-r1-0528	0.523	0.127
mistralai/mistral-medium-3.1	0.514	0.078
xiaomi/mimo-v2.5-pro	0.501	0.122
openai/gpt-5.4-mini	0.488	0.042
mistralai/codestral-2508	0.487	0.049
meituan/longcat-2.0	0.482	0.109
qwen/qwen3.7-plus	0.459	0.017
cohere/command-a	0.455	0.000
nvidia/nemotron-3-ultra-550b-a55b	0.380	0.117
stepfun/step-3.7-flash	0.366	0.154
mistralai/mistral-large-2512	0.333	0.008
thinkingmachines/inkling	0.315	0.112
deepseek/deepseek-v3.2	0.303	0.317
openai/gpt-3.5-turbo	0.286	0.000
microsoft/phi-4	0.228	0.085
thinkingmachines/inkling-small	0.219	0.212
openai/o4-mini	0.202	0.211
meta-llama/llama-3.1-8b-instruct	0.124	0.170
qwen/qwen3.8-max	0.050	0.112
mistralai/mistral-small-2603	0.000	0.000
qwen/qwen3-8b	0.000	0.000

ep-the-brilliant-idiots-0bb9bf634c8e: Class Rank

• Podcast: the-brilliant-idiots
• Duration: 119.9 min
• Truth ads: 3

Model	F1	F1 stdev
claude-haiku-4-5-20251001	1.000	0.000
claude-sonnet-4-6	1.000	0.000
google/gemini-3.5-flash	1.000	0.000
google/gemini-3.6-flash	1.000	0.000
deepseek/deepseek-v4-flash-0731	0.943	0.078
openai/gpt-5.5	0.943	0.078
deepseek/deepseek-v4-flash	0.903	0.092
qwen/qwen3.6-plus	0.886	0.186
moonshotai/kimi-k3	0.864	0.089
google/gemini-3.1-pro-preview	0.857	0.000
x-ai/grok-4.5	0.857	0.000
tencent/hy3	0.836	0.120
openai/gpt-5.6-sol	0.834	0.145
openai/gpt-oss-20b	0.825	0.120
qwen/qwen3.7-max	0.821	0.110
qwen/qwen3.6-flash	0.810	0.230
qwen/qwen3.5-plus-02-15	0.800	0.112
openai/gpt-5.6-luna	0.793	0.059
openai/gpt-5.6-terra	0.791	0.148
qwen/qwen3.7-flash	0.769	0.167
openai/gpt-oss-120b	0.762	0.135
nvidia/nemotron-3-super-120b-a12b	0.754	0.102
x-ai/grok-4.3	0.743	0.156
google/gemini-3.5-flash-lite	0.700	0.046
google/gemma-4-26b-a4b-it	0.697	0.115
mistralai/mistral-medium-3.1	0.681	0.074
meta/muse-spark-1.1	0.680	0.164
openai/gpt-5.4	0.670	0.053
claude-fable-5	0.667	0.000
claude-opus-5	0.667	0.000
meta-llama/llama-4-maverick	0.667	0.000
mistralai/mistral-medium-3-5	0.667	0.000
xiaomi/mimo-v2.5-pro	0.641	0.151
google/gemini-2.5-pro	0.633	0.075
microsoft/phi-4	0.620	0.164
minimax/minimax-m3	0.599	0.299
deepseek/deepseek-v4-pro	0.597	0.234
qwen/qwen3.5-27b	0.586	0.190
claude-opus-4-7	0.578	0.122
meta-llama/llama-3.3-70b-instruct	0.571	0.000
claude-opus-4-8	0.556	0.104
google/gemma-4-31b-it	0.550	0.112
google/gemini-3.1-flash-lite	0.540	0.055
z-ai/glm-5.2	0.527	0.120
claude-sonnet-5	0.503	0.182
deepseek/deepseek-v3.2	0.500	0.000
google/gemini-2.5-flash	0.500	0.000
thinkingmachines/inkling-small	0.483	0.149
deepseek/deepseek-r1-0528	0.472	0.158
qwen/qwen3-14b	0.470	0.141
deepseek/deepseek-r1	0.462	0.134
cohere/command-r-plus-08-2024	0.460	0.055
qwen/qwen3.8-max	0.460	0.055
thinkingmachines/inkling	0.460	0.456
openai/o3	0.451	0.306
meta-llama/llama-4-scout	0.447	0.274
google/gemini-2.5-flash-lite	0.438	0.069
moonshotai/kimi-k2.6	0.397	0.289
deepseek/deepseek-r1-distill-llama-70b	0.374	0.167
openai/o4-mini	0.360	0.351
qwen/qwen3-235b-a22b-2507	0.359	0.061
inclusionai/ring-2.6-1t	0.350	0.241
xiaomi/mimo-v2.5	0.341	0.181
cohere/command-a	0.335	0.102
mistralai/codestral-2508	0.300	0.021
meituan/longcat-2.0	0.280	0.284
qwen/qwen3.7-plus	0.270	0.009
nvidia/nemotron-3-ultra-550b-a55b	0.244	0.250
openai/gpt-5.4-mini	0.241	0.008
mistralai/mistral-large-2512	0.239	0.028
openai/gpt-3.5-turbo	0.211	0.008
meta-llama/llama-3.1-8b-instruct	0.090	0.083
mistralai/mistral-small-2603	0.000	0.000
qwen/qwen3-8b	0.000	0.000

ep-the-tim-dillon-show-f62bd5fa1cfe: 495 - Hantavirus Cruise & iPad Babies

• Podcast: the-tim-dillon-show
• Duration: 80.1 min
• Truth ads: 6

Model	F1	F1 stdev
qwen/qwen3.7-max	0.848	0.034
x-ai/grok-4.3	0.842	0.040
google/gemini-2.5-flash	0.833	0.000
google/gemini-3.1-flash-lite	0.833	0.000
mistralai/mistral-medium-3.1	0.833	0.000
google/gemini-3.5-flash-lite	0.825	0.070
claude-haiku-4-5-20251001	0.812	0.047
claude-fable-5	0.800	0.000
claude-opus-4-8	0.800	0.000
claude-opus-5	0.800	0.000
deepseek/deepseek-v4-pro	0.800	0.000
google/gemini-2.5-pro	0.800	0.000
google/gemini-3.1-pro-preview	0.800	0.000
google/gemini-3.5-flash	0.800	0.000
google/gemini-3.6-flash	0.800	0.000
moonshotai/kimi-k3	0.800	0.000
openai/gpt-5.5	0.800	0.000
openai/gpt-5.6-luna	0.800	0.000
openai/gpt-5.6-terra	0.800	0.000
x-ai/grok-4.5	0.800	0.000
claude-sonnet-5	0.785	0.033
openai/gpt-5.4	0.785	0.033
qwen/qwen3.7-flash	0.785	0.033
openai/gpt-5.6-sol	0.773	0.060
qwen/qwen3.5-plus-02-15	0.748	0.047
deepseek/deepseek-v4-flash	0.733	0.094
claude-opus-4-7	0.727	0.000
qwen/qwen3.6-flash	0.720	0.110
mistralai/mistral-medium-3-5	0.716	0.063
qwen/qwen3.6-plus	0.691	0.081
thinkingmachines/inkling-small	0.691	0.088
z-ai/glm-5.2	0.678	0.053
google/gemma-4-31b-it	0.675	0.240
claude-sonnet-4-6	0.667	0.000
openai/gpt-oss-120b	0.662	0.114
qwen/qwen3.8-max	0.646	0.117
deepseek/deepseek-v4-flash-0731	0.636	0.129
deepseek/deepseek-r1	0.620	0.117
meta/muse-spark-1.1	0.604	0.192
qwen/qwen3-235b-a22b-2507	0.597	0.126
qwen/qwen3.7-plus	0.580	0.020
xiaomi/mimo-v2.5	0.578	0.121
deepseek/deepseek-r1-0528	0.553	0.200
google/gemma-4-26b-a4b-it	0.545	0.201
openai/gpt-5.4-mini	0.514	0.132
openai/gpt-oss-20b	0.483	0.167
deepseek/deepseek-v3.2	0.476	0.070
nvidia/nemotron-3-ultra-550b-a55b	0.463	0.168
nvidia/nemotron-3-super-120b-a12b	0.462	0.155
google/gemini-2.5-flash-lite	0.459	0.069
mistralai/mistral-large-2512	0.449	0.048
openai/o3	0.425	0.222
microsoft/phi-4	0.423	0.076
mistralai/codestral-2508	0.401	0.027
minimax/minimax-m3	0.399	0.367
moonshotai/kimi-k2.6	0.394	0.167
xiaomi/mimo-v2.5-pro	0.386	0.252
qwen/qwen3.5-27b	0.373	0.147
meta-llama/llama-4-maverick	0.364	0.000
deepseek/deepseek-r1-distill-llama-70b	0.346	0.107
tencent/hy3	0.346	0.118
meta-llama/llama-3.3-70b-instruct	0.329	0.097
stepfun/step-3.7-flash	0.279	0.016
meta-llama/llama-4-scout	0.256	0.144
cohere/command-r-plus-08-2024	0.239	0.177
cohere/command-a	0.218	0.081
thinkingmachines/inkling	0.202	0.211
openai/gpt-3.5-turbo	0.200	0.068
qwen/qwen3-14b	0.185	0.182
meituan/longcat-2.0	0.152	0.140
meta-llama/llama-3.1-8b-instruct	0.140	0.097
inclusionai/ring-2.6-1t	0.094	0.130
mistralai/mistral-small-2603	0.000	0.000
openai/o4-mini	0.000	0.000
qwen/qwen3-8b	0.000	0.000

ep-tosh-show-5f6894439bb6: My Mom - Emergency Pod

• Podcast: tosh-show
• Duration: 41.4 min
• Truth ads: 5

Model	F1	F1 stdev
openai/gpt-5.6-sol	0.889	0.000
claude-sonnet-4-6	0.879	0.125
claude-fable-5	0.864	0.041
google/gemini-3.6-flash	0.857	0.053
x-ai/grok-4.5	0.848	0.034
claude-opus-4-8	0.822	0.049
meta-llama/llama-4-scout	0.787	0.142
x-ai/grok-4.3	0.745	0.145
openai/gpt-oss-120b	0.743	0.153
qwen/qwen3.7-plus	0.728	0.124
google/gemini-3.1-flash-lite	0.727	0.000
google/gemini-3.5-flash-lite	0.727	0.000
mistralai/codestral-2508	0.727	0.000
mistralai/mistral-medium-3.1	0.727	0.000
qwen/qwen3.5-plus-02-15	0.727	0.000
qwen/qwen3.6-flash	0.723	0.122
minimax/minimax-m3	0.719	0.066
qwen/qwen3.7-flash	0.718	0.055
mistralai/mistral-medium-3-5	0.715	0.027
openai/gpt-5.6-terra	0.711	0.099
qwen/qwen3.7-max	0.681	0.047
google/gemini-3.1-pro-preview	0.676	0.070
z-ai/glm-5.2	0.670	0.115
deepseek/deepseek-r1	0.669	0.147
claude-haiku-4-5-20251001	0.669	0.040
mistralai/mistral-large-2512	0.667	0.000
openai/gpt-5.6-luna	0.667	0.157
claude-opus-5	0.667	0.000
google/gemini-3.5-flash	0.656	0.023
claude-sonnet-5	0.647	0.104
openai/gpt-5.5	0.622	0.099
google/gemini-2.5-flash-lite	0.609	0.080
cohere/command-a	0.600	0.000
moonshotai/kimi-k3	0.599	0.248
qwen/qwen3.6-plus	0.579	0.363
google/gemini-2.5-flash	0.571	0.000
xiaomi/mimo-v2.5	0.569	0.179
deepseek/deepseek-r1-0528	0.549	0.120
openai/gpt-5.4-mini	0.542	0.114
tencent/hy3	0.539	0.246
claude-opus-4-7	0.538	0.215
qwen/qwen3-235b-a22b-2507	0.529	0.172
deepseek/deepseek-v3.2	0.527	0.078
openai/gpt-5.4	0.520	0.110
moonshotai/kimi-k2.6	0.510	0.103
google/gemma-4-31b-it	0.476	0.145
meituan/longcat-2.0	0.473	0.093
nvidia/nemotron-3-ultra-550b-a55b	0.471	0.325
deepseek/deepseek-v4-pro	0.456	0.273
meta/muse-spark-1.1	0.451	0.162
meta-llama/llama-4-maverick	0.444	0.000
google/gemini-2.5-pro	0.432	0.166
openai/gpt-oss-20b	0.375	0.275
meta-llama/llama-3.3-70b-instruct	0.363	0.065
google/gemma-4-26b-a4b-it	0.358	0.175
deepseek/deepseek-v4-flash	0.355	0.156
qwen/qwen3.5-27b	0.350	0.267
nvidia/nemotron-3-super-120b-a12b	0.319	0.201
cohere/command-r-plus-08-2024	0.295	0.021
qwen/qwen3-14b	0.278	0.094
xiaomi/mimo-v2.5-pro	0.253	0.258
mistralai/mistral-small-2603	0.222	0.000
openai/gpt-3.5-turbo	0.218	0.010
microsoft/phi-4	0.217	0.046
meta-llama/llama-3.1-8b-instruct	0.216	0.032
deepseek/deepseek-v4-flash-0731	0.203	0.196
thinkingmachines/inkling-small	0.181	0.262
openai/o3	0.178	0.169
openai/o4-mini	0.133	0.183
inclusionai/ring-2.6-1t	0.129	0.197
stepfun/step-3.7-flash	0.067	0.149
deepseek/deepseek-r1-distill-llama-70b	0.050	0.112
qwen/qwen3-8b	0.000	0.000
qwen/qwen3.8-max	0.000	0.000
thinkingmachines/inkling	0.000	0.000

Parser stress test

How each model’s responses were actually parsed. Columns are extraction methods, ordered alphabetically; rows are models, sorted by parse-failure rate (cleanest at top). json_array_direct is the happy path: a bare JSON array we could json.loads and process immediately. markdown_code_block means we had to strip triple-backtick fences first; json_object_* means the model wrapped the array in an outer object and we had to find the array key; regex_* are last-resort recovery paths. A model that needs anything but json_array_direct for most calls is fragile. It works today, but a small prompt change can break the parser.

Model	bracket_fallback	json_array_direct	json_object_ad_key	json_object_ads_detected_key	json_object_ads_key	json_object_advertisement_segments_key	json_object_no_ads	json_object_segments_key	json_object_single_ad	json_object_single_ad_truncated	markdown_code_block	parse_failure	regex_json_array
claude-fable-5	0	855	0	0	0	0	0	0	0	0	0	0	0
claude-haiku-4-5-20251001	0	855	0	0	0	0	0	0	0	0	0	0	0
claude-opus-4-7	0	855	0	0	0	0	0	0	0	0	0	0	0
claude-opus-4-8	0	855	0	0	0	0	0	0	0	0	0	0	0
claude-opus-5	0	855	0	0	0	0	0	0	0	0	0	0	0
claude-sonnet-4-6	0	855	0	0	0	0	0	0	0	0	0	0	0
claude-sonnet-5	0	854	0	0	0	0	0	0	1	0	0	0	0
cohere/command-a	0	0	0	0	0	0	21	0	834	0	0	0	0
cohere/command-r-plus-08-2024	0	0	0	0	5	0	661	0	189	0	0	0	0
deepseek/deepseek-v3.2	0	853	0	0	2	0	0	0	0	0	0	0	0
google/gemini-2.5-flash	0	852	0	0	0	0	0	0	0	3	0	0	0
google/gemini-3.1-flash-lite	0	771	0	0	0	0	0	0	0	1	0	0	83
google/gemini-3.1-pro-preview	0	855	0	0	0	0	0	0	0	0	0	0	0
google/gemini-3.5-flash	0	852	0	0	0	0	0	0	0	0	0	0	3
google/gemini-3.5-flash-lite	0	853	0	0	0	0	0	0	0	1	0	0	1
google/gemini-3.6-flash	0	854	0	0	0	0	0	0	0	1	0	0	0
google/gemma-4-26b-a4b-it	0	0	0	0	73	0	372	0	410	0	0	0	0
google/gemma-4-31b-it	0	14	0	0	378	0	283	0	180	0	0	0	0
meta-llama/llama-3.1-8b-instruct	0	208	0	0	0	0	138	0	509	0	0	0	0
meta-llama/llama-3.3-70b-instruct	0	147	0	0	0	0	380	0	322	4	0	0	2
meta-llama/llama-4-maverick	0	0	0	0	0	0	259	0	596	0	0	0	0
mistralai/codestral-2508	0	855	0	0	0	0	0	0	0	0	0	0	0
mistralai/mistral-large-2512	0	855	0	0	0	0	0	0	0	0	0	0	0
mistralai/mistral-medium-3-5	0	855	0	0	0	0	0	0	0	0	0	0	0
mistralai/mistral-medium-3.1	0	855	0	0	0	0	0	0	0	0	0	0	0
mistralai/mistral-small-2603	0	855	0	0	0	0	0	0	0	0	0	0	0
openai/gpt-5.4	0	0	0	0	0	0	274	0	581	0	0	0	0
openai/gpt-5.4-mini	0	0	0	0	0	0	218	0	637	0	0	0	0
openai/gpt-5.6-sol	0	0	0	0	53	0	368	3	431	0	0	0	0
openai/gpt-5.6-terra	0	0	0	0	0	0	458	0	397	0	0	0	0
qwen/qwen3-235b-a22b-2507	0	227	0	0	0	0	169	0	459	0	0	0	0
x-ai/grok-4.3	0	855	0	0	0	0	0	0	0	0	0	0	0
x-ai/grok-4.5	0	855	0	0	0	0	0	0	0	0	0	0	0
nvidia/nemotron-3-ultra-550b-a55b	0	448	0	0	59	0	56	37	252	0	0	3	0
openai/gpt-5.5	0	0	0	0	0	0	489	0	363	0	0	3	0
meta/muse-spark-1.1	0	0	0	0	28	0	643	3	175	2	0	4	0
minimax/minimax-m3	0	340	0	0	4	0	262	0	48	0	188	4	9
qwen/qwen3.5-plus-02-15	0	851	0	0	0	0	0	0	0	0	0	4	0
qwen/qwen3.7-max	0	833	0	0	0	0	0	0	8	10	0	4	0
openai/gpt-3.5-turbo	0	0	0	0	0	0	19	0	831	0	0	5	0
openai/o3	0	0	0	0	87	0	613	30	119	0	0	6	0
openai/gpt-5.6-luna	0	0	0	0	32	0	368	0	446	0	0	9	0
openai/gpt-oss-120b	0	28	0	0	181	0	429	0	202	1	0	9	5
google/gemini-2.5-pro	0	805	0	0	0	0	0	0	0	9	0	12	29
microsoft/phi-4	0	828	0	0	0	0	4	0	3	7	0	13	0
google/gemini-2.5-flash-lite	0	791	0	0	0	0	0	0	0	47	0	17	0
qwen/qwen3.7-plus	0	242	0	0	3	0	86	0	500	2	0	22	0
xiaomi/mimo-v2.5-pro	0	0	0	1	133	7	552	11	119	5	0	22	5
xiaomi/mimo-v2.5	0	0	0	1	54	0	160	2	492	48	18	26	54
meta-llama/llama-4-scout	31	56	0	0	349	0	262	10	110	3	0	27	7
z-ai/glm-5.2	0	5	4	0	95	0	86	0	635	1	0	29	0
qwen/qwen3.6-flash	0	820	0	0	0	0	0	0	0	5	0	30	0
qwen/qwen3.7-flash	0	810	0	0	0	0	0	0	0	12	0	33	0
deepseek/deepseek-v4-pro	0	434	0	0	10	0	39	205	118	3	0	45	1
deepseek/deepseek-v4-flash	0	93	0	0	175	0	266	6	268	0	0	47	0
moonshotai/kimi-k3	0	361	0	0	181	0	27	0	6	122	0	50	108
deepseek/deepseek-r1	0	494	0	0	182	0	0	1	90	3	0	85	0
qwen/qwen3.6-plus	0	748	0	0	0	0	0	0	0	21	0	86	0
deepseek/deepseek-r1-0528	0	535	0	0	154	0	7	4	55	4	0	96	0
openai/gpt-oss-20b	0	26	0	0	121	0	430	0	176	3	0	99	0
inclusionai/ring-2.6-1t	1	0	0	0	0	0	568	6	168	2	0	110	0
nvidia/nemotron-3-super-120b-a12b	0	520	0	0	1	0	40	0	160	6	0	127	1
qwen/qwen3-14b	0	0	0	0	0	0	10	0	712	2	0	131	0
deepseek/deepseek-r1-distill-llama-70b	0	36	0	0	0	0	0	0	0	1	110	135	573
deepseek/deepseek-v4-flash-0731	0	224	0	0	9	0	222	0	238	3	0	158	1
qwen/qwen3.8-max	0	621	0	0	0	0	0	0	3	1	0	230	0
tencent/hy3	12	144	0	0	10	0	198	0	249	6	0	236	0
thinkingmachines/inkling-small	0	0	0	0	0	0	281	0	284	4	0	250	36
qwen/qwen3.5-27b	0	460	0	0	0	0	124	0	17	1	0	253	0
moonshotai/kimi-k2.6	0	270	0	0	57	0	77	3	188	6	0	254	0
stepfun/step-3.7-flash	0	0	0	0	380	0	4	0	103	4	0	234	0
meituan/longcat-2.0	0	0	0	0	414	0	8	0	1	2	0	429	1
thinkingmachines/inkling	81	0	0	0	3	0	142	0	188	2	0	430	0
qwen/qwen3-8b	15	1	0	0	0	0	79	0	0	0	0	760	0
openai/o4-mini	0	0	0	0	0	0	24	1	52	0	0	778	0

Methodology

Reproducibility settings used for this run. The benchmark sends the same prompts MinusPod sends in production (same system prompt, same sponsor list, same windowing) so the F1 numbers here are directly relevant to production accuracy decisions. Cost is recomputed at report time from token counts against the active pricing snapshot, so all rows compare at the same prices regardless of when the actual call ran.

• Trials per (model, episode): 5, temperature 0.0
• max_tokens: 4096 (matches MinusPod production)
• response_format: json_object (with prompt-injection fallback when provider rejects native)
• Window size: 10 min, overlap: 3 min (imported from MinusPod’s create_windows)
• Pricing snapshot: 2026-08-07T17:23:44.665500Z
• Corpus episodes: 14

Transcript source

segments.json for every corpus episode is pulled byte-exact from the source MinusPod instance’s original-segments endpoint. The transcript itself was generated by faster-whisper inside that instance, not by the benchmark. Model choice and decoding params affect what gets transcribed, which sets an upper bound on what every benchmarked LLM can find.

Whisper config:

Setting	Value
Model	large-v3
Backend	local (faster-whisper, CUDA GPU)
Compute type	auto (resolves to float16 on CUDA)
Language	en (forced English, not auto-detect)
VAD gap detection	on (start 3.0s / mid 8.0s / tail 3.0s)

model.transcribe() invocation (from src/transcriber.py):

WhisperModel(model_size="large-v3", device="cuda", compute_type="auto")
model.transcribe(
    audio,
    language="en",
    initial_prompt=<podcast name + SEED_SPONSORS vocabulary>,
    beam_size=5,
    batch_size=<adaptive: 16/12/8/4 by episode length>,
    word_timestamps=True,
    vad_filter=True,
    vad_parameters={"min_silence_duration_ms": 1000, "speech_pad_ms": 600, "threshold": 0.3},
)

The initial_prompt carries a sponsor vocabulary so Whisper produces consistent spellings (Athletic Greens rather than AG1, ExpressVPN rather than express vpn). This biases what shows up in the transcript and therefore what every benchmarked LLM is scored against.

Sponsor vocabulary (254 canonical sponsors, 44 of them with explicit alias spellings totaling 48 aliases; from src/utils/constants.py SEED_SPONSORS). Laid out in two side-by-side groups, read top-to-bottom in each group.

Sponsor	Aliases	Category	Sponsor	Aliases	Category
1Password	One Password	tech	MacPaw	CleanMyMac	tech
Acorns	-	finance	Magic Mind	-	beverage
ADT	-	home	Magic Spoon	-	food
Affirm	-	finance_fintech	Mailchimp	-	tech_software_saas
Airbnb	-	travel_hospitality	Manscaped	-	personal
Airtable	-	tech_software_saas	MasterClass	Master Class	education
Alani Nu	-	food_beverage_nutrition	McDonald’s	-	food_beverage_nutrition
Allbirds	-	ecommerce_retail_dtc	Mercury	-	finance_fintech
Alo Yoga	-	ecommerce_retail_dtc	Meter	-	b2b_startup
Amazon	-	retail	Midjourney	-	tech_software_saas
Anthropic	-	tech_software_saas	Mint Mobile	MintMobile	telecom
Apple TV+	-	media_streaming	Miro	-	tech
Asana	-	tech_software_saas	Momentous	-	mental_health_wellness
AT&T	-	telecom	Monarch Money	-	finance
Athletic Brewing	-	beverage	Monday.com	Monday	tech
Athletic Greens	AG1, AG One	health	Native	-	personal
Audible	-	entertainment	NerdWallet	-	finance_fintech
Aura	-	tech	Netflix	-	media_streaming
Babbel	-	education	NetSuite	Net Suite	tech
BetMGM	Bet MGM	gambling	Noom	-	mental_health_wellness
BetterHelp	Better Help	health	NordVPN	Nord VPN	vpn
Betterment	-	finance	Notion	-	tech
Bill.com	-	finance_fintech	Nutrafol	-	health
Birchbox	-	ecommerce_retail_dtc	Okta	-	tech_software_saas
Bitwarden	Bit Warden	tech	OLIPOP	-	food_beverage_nutrition
Blinkist	-	education	OneSkin	One Skin	personal
Bloom Nutrition	-	food_beverage_nutrition	OpenAI	-	tech_software_saas
Blue Apron	-	food	Outdoor Voices	-	ecommerce_retail_dtc
Bombas	-	apparel	OutSystems	-	tech
Booking.com	-	travel_hospitality	PagerDuty	-	b2b_startup
Bose	-	electronics	Paramount+	-	media_streaming
Brex	-	finance_fintech	Patreon	-	tech_software_saas
Brilliant	-	tech_software_saas	Perplexity	-	tech_software_saas
Brooklinen	-	home	Plaid	-	finance_fintech
Butcher Box	ButcherBox	food	PolicyGenius	Policy Genius	finance
CacheFly	-	tech	Poppi	-	food_beverage_nutrition
Caesars Sportsbook	-	gaming_sports_betting	Poshmark	-	ecommerce_retail_dtc
Calm	-	health	Progressive	-	finance
Canva	-	tech	Public.com	-	finance_fintech
Capital One	-	finance	Pura	-	home_security
Care/of	Care of, Careof	health	Purple	-	home
CarMax	Car Max	auto	QuickBooks	-	finance_fintech
Carvana	-	auto	Quince	-	apparel
Casper	-	home	Quip	-	personal
Cerebral	-	mental_health_wellness	Ramp	-	finance_fintech
Chime	-	finance_fintech	Raycon	-	electronics
ClickUp	-	tech_software_saas	Retool	-	tech_software_saas
Cloudflare	-	tech_software_saas	Ring	-	home
Coinbase	-	finance_fintech	Rippling	-	b2b_startup
Comcast	-	telecom	Ritual	-	health
Cozy Earth	-	home	Ro	-	mental_health_wellness
Credit Karma	-	finance	Robinhood	-	finance_fintech
CrowdStrike	-	tech_software_saas	Rocket Lawyer	-	insurance_legal
Cursor	-	tech_software_saas	Rocket Money	RocketMoney, Truebill	finance
Databricks	-	tech_software_saas	Roman	-	health
Datadog	-	tech_software_saas	Rosetta Stone	-	education
Deel	-	business	Rothy’s	-	ecommerce_retail_dtc
DeleteMe	Delete Me	tech	Saatva	-	ecommerce_retail_dtc
Disney+	-	media_streaming	Salesforce	-	tech_software_saas
DocuSign	-	tech_software_saas	SeatGeek	-	gaming_sports_betting
Dollar Shave Club	DSC	personal	Seed	-	health
DoorDash	Door Dash	food	SendGrid	-	tech_software_saas
DraftKings	Draft Kings	gambling	ServiceNow	-	tech_software_saas
Duolingo	-	tech_software_saas	Shein	-	ecommerce_retail_dtc
eBay Motors	-	auto	Shopify	-	tech
Eight Sleep	-	mental_health_wellness	SimpliSafe	Simpli Safe	home
ElevenLabs	-	tech_software_saas	SiriusXM	-	media_streaming
ESPN Bet	-	gaming_sports_betting	Skillshare	-	tech_software_saas
Everlane	-	ecommerce_retail_dtc	SKIMS	-	ecommerce_retail_dtc
EveryPlate	-	food_beverage_nutrition	Skyscanner	-	travel_hospitality
Expedia	-	travel_hospitality	Slack	-	tech_software_saas
ExpressVPN	Express VPN	vpn	Snowflake	-	tech_software_saas
FabFitFun	-	ecommerce_retail_dtc	SoFi	-	finance
Factor	-	food	Spaceship	-	tech
FanDuel	Fan Duel	gambling	Splunk	-	b2b_startup
Figma	-	tech_software_saas	Spotify	-	media_streaming
Ford	-	auto	Squarespace	Square Space	tech
Framer	-	tech	Stamps.com	Stamps	business
FreshBooks	-	finance_fintech	Starbucks	-	food_beverage_nutrition
Function Health	-	mental_health_wellness	State Farm	-	finance
Function of Beauty	-	personal	Stitch Fix	-	ecommerce_retail_dtc
Gametime	Game Time	entertainment	StockX	-	ecommerce_retail_dtc
Geico	-	finance	Stripe	-	finance_fintech
GitHub	-	tech_software_saas	StubHub	-	gaming_sports_betting
GitHub Copilot	-	tech_software_saas	Substack	-	tech_software_saas
GOAT	-	ecommerce_retail_dtc	T-Mobile	TMobile	telecom
GoodRx	Good Rx	health	Talkspace	-	mental_health_wellness
Gopuff	-	ecommerce_retail_dtc	Temu	-	ecommerce_retail_dtc
Grammarly	-	tech	Ten Thousand	-	ecommerce_retail_dtc
Green Chef	GreenChef	food	Thinkst Canary	-	tech
Grubhub	Grub Hub	food	Thorne	-	mental_health_wellness
Gusto	-	b2b_startup	ThreatLocker	-	tech
Harry’s	Harrys	personal	ThredUp	-	ecommerce_retail_dtc
HBO Max	-	media_streaming	Thrive Market	-	food
Headspace	Head Space	health	Toyota	-	auto
Helix Sleep	Helix	home	Transparent Labs	-	food_beverage_nutrition
HelloFresh	Hello Fresh	food	Turo	-	automotive_transport
Hers	-	health	Twilio	-	tech_software_saas
Hims	-	health	Uber	-	automotive_transport
Honeylove	Honey Love	apparel	Uber Eats	UberEats	food
Hopper	-	travel_hospitality	UnitedHealth Group	-	finance_fintech
HubSpot	Hub Spot	tech	Vanta	-	tech
Huel	-	food_beverage_nutrition	Veeam	-	tech
Hyundai	-	auto	Vercel	-	tech_software_saas
iHeartRadio	-	media_streaming	Verizon	-	telecom
Imperfect Foods	-	food_beverage_nutrition	Visible	-	telecom
Incogni	-	tech	Vrbo	-	travel_hospitality
Indeed	-	jobs	Vuori	-	ecommerce_retail_dtc
Inside Tracker	-	mental_health_wellness	Warby Parker	-	ecommerce_retail_dtc
Instacart	-	food	Wayfair	-	ecommerce_retail_dtc
Intuit	-	finance_fintech	Waymo	-	automotive_transport
Joovv	-	mental_health_wellness	Wealthfront	-	finance
Kayak	-	travel_hospitality	WebBank	-	finance_fintech
Klarna	-	finance_fintech	Webflow	-	b2b_startup
Klaviyo	-	tech_software_saas	WhatsApp	-	tech
LegalZoom	-	insurance_legal	WHOOP	-	mental_health_wellness
Lemonade	-	finance	Workday	-	tech_software_saas
Levels	-	mental_health_wellness	Xero	-	finance_fintech
Liberty Mutual	-	finance	YouTube	-	media_streaming
Lime	-	automotive_transport	YouTube TV	-	media_streaming
Linear	-	tech_software_saas	Zapier	-	tech
LinkedIn	LinkedIn Jobs	jobs	Zendesk	-	tech_software_saas
Liquid IV	Liquid I.V.	health	ZipRecruiter	Zip Recruiter	jobs
LMNT	Element	health	ZocDoc	Zoc Doc	health
Loom	-	tech_software_saas	Zoom	-	tech_software_saas
Lululemon	-	ecommerce_retail_dtc	Zscaler	-	tech
Lyft	-	automotive_transport	Zyn	ZYN, Zinn	tobacco_nicotine

Mishearing corrections (174 entries, from src/utils/constants.py SPONSOR_ALIASES). Applied post-transcription to normalize Whisper output toward the canonical sponsor name. Distinct from the aliases column above, which lists intentional alternative spellings (e.g. AG1 vs Athletic Greens); the entries below are mostly Whisper mishearings (e.g. a firm -> Affirm, xerox -> Xero). Laid out in three side-by-side groups, read top-to-bottom in each group.

Heard as	Normalized to	Heard as	Normalized to	Heard as	Normalized to
1 password	1Password	good-rx	GoodRx	patron	Patreon
8 sleep	Eight Sleep	green chef	Green Chef	pay tree on	Patreon
8-sleep	Eight Sleep	green-chef	Green Chef	perplexity ai	Perplexity
a firm	Affirm	greenchef	Green Chef	perplexity-ai	Perplexity
a g one	Athletic Greens	grub hub	Grubhub	policy genius	PolicyGenius
ag 1	Athletic Greens	grub-hub	Grubhub	policy-genius	PolicyGenius
ag one	Athletic Greens	harrys	Harry’s	pyura	Pura
ag1	Athletic Greens	head space	Headspace	ray con	Raycon
athlean x	Athlean-X	head-space	Headspace	ray-con	Raycon
athlean-x	Athlean-X	hello fresh	HelloFresh	re tool	Retool
athletic greens one	Athletic Greens	hello-fresh	HelloFresh	ro gain	Rogaine
athleticgreens	Athletic Greens	him's	Hims	ro-gaine	Rogaine
bet mgm	BetMGM	hims & hers	Hims & Hers	rocket money	Rocket Money
bet-mgm	BetMGM	hims and hers	Hims & Hers	rocket-money	Rocket Money
better help	BetterHelp	honey love	Honeylove	rocketlawyer	Rocket Lawyer
better-help	BetterHelp	honey-love	Honeylove	rocketmoney	Rocket Money
birch box	Birchbox	honeylove	Honeylove	rocketmortgage	Rocket Mortgage
birch-box	Birchbox	hub spot	HubSpot	seat geek	SeatGeek
bit warden	Bitwarden	hub-spot	HubSpot	seat-geek	SeatGeek
bit-warden	Bitwarden	hubs pot	HubSpot	shop a fly	Shopify
blueapron	Blue Apron	imperfect foods	Imperfect Foods	shop fly	Shopify
brecks	Brex	imperfectfoods	Imperfect Foods	shop ify	Shopify
butcher box	Butcher Box	insta cart	Instacart	simpli safe	SimpliSafe
butcher-box	Butcher Box	insta-cart	Instacart	simpli-safe	SimpliSafe
butcherbox	Butcher Box	l m n t	LMNT	simply safe	SimpliSafe
car max	CarMax	legal zoom	LegalZoom	sky scanner	Skyscanner
car-max	CarMax	legal-zoom	LegalZoom	sky-scanner	Skyscanner
cloud flare	Cloudflare	legalzoom	LegalZoom	so fi	SoFi
cloud-flare	Cloudflare	liquid i v	Liquid IV	so-fi	SoFi
co pilot	GitHub Copilot	liquid i.v.	Liquid IV	square space	Squarespace
co-pilot	GitHub Copilot	liquid iv	Liquid IV	square-space	Squarespace
copilot	GitHub Copilot	liquidiv	Liquid IV	stamp dot com	Stamps.com
creditkarma	Credit Karma	magic mind	Magic Mind	stitch fix	Stitch Fix
delete me	DeleteMe	magic spoon	Magic Spoon	stitch-fix	Stitch Fix
delete-me	DeleteMe	magicmind	Magic Mind	stitchfix	Stitch Fix
dollarshaveclub	Dollar Shave Club	magicspoon	Magic Spoon	stub hub	StubHub
door dash	DoorDash	master class	MasterClass	stub-hub	StubHub
door-dash	DoorDash	master-class	MasterClass	sub stack	Substack
draft kings	DraftKings	mercury bank	Mercury	sub-stack	Substack
draft-kings	DraftKings	mercury-bank	Mercury	thrive market	Thrive Market
eight-sleep	Eight Sleep	mint mobile	Mint Mobile	thrivemarket	Thrive Market
eightsleep	Eight Sleep	mint-mobile	Mint Mobile	transparent labs	Transparent Labs
element	LMNT	mintmobile	Mint Mobile	transparentlabs	Transparent Labs
every plate	EveryPlate	monarch money	Monarch Money	uber eats	Uber Eats
every-plate	EveryPlate	monarch-money	Monarch Money	uber-eats	Uber Eats
express vpn	ExpressVPN	monarchmoney	Monarch Money	ubereats	Uber Eats
express-vpn	ExpressVPN	my protein	Myprotein	ver cell	Vercel
fab fit fun	FabFitFun	my ro	Miro	ver sel	Vercel
fab-fit-fun	FabFitFun	myprotein	Myprotein	wealth front	Wealthfront
fan duel	FanDuel	net suite	NetSuite	wealth-front	Wealthfront
fan-duel	FanDuel	net-suite	NetSuite	woop	Whoop
game time	Gametime	nord vpn	NordVPN	xerox	Xero
game-time	Gametime	nord-vpn	NordVPN	zero	Xero
gametime	Gametime	one password	1Password	zip recruiter	ZipRecruiter
github-copilot	GitHub Copilot	one skin	OneSkin	zip-recruiter	ZipRecruiter
go puff	Gopuff	one-password	1Password	zoc doc	ZocDoc
go-puff	Gopuff	one-skin	OneSkin	zoc-doc	ZocDoc
good rx	GoodRx	p ninety x	P90X	zock doc	ZocDoc

Run Metadata

• Report generated: 2026-08-08T18:52:53Z
• Unique work units (current state, last-write-wins after retries): 64125
• Raw rows in calls.jsonl: 70072 (5947 superseded by later retries; kept for audit)
• Successful: 63986
• Failed: 139
• Lifetime list-price cost (sum of at-runtime costs, includes superseded rows): $632.0457
• Lifetime tokens (same basis): 410,609,818 in + 53,769,250 out = 464,379,068
• Note: every input token is priced at list rate. Providers that serve a repeated prompt from cache bill less than this, and the harness does not record cache hits, so a real invoice for this run will come in under the figure above.
• Active pricing snapshot: 2026-08-07T17:23:44.665500Z
• System prompt: live (sha256:1030b29e)