Vision & documents

Video-MME (with subtitle) leaderboard

Name: benchgap: LLM benchmark scores, measured and estimated
Creator: benchgap

As of 2026-10-07, the highest measured score on Video-MME (with subtitle) is 90.4% by Qwen3.8 Max. 52 more models have estimated scores, calibrated from the benchmarks they were measured on.

Measured scores: benchlm.ai.

#	Model	Score	Source
1	Qwen3.8 Max	90.4%	measured
2	Claude Mythos 5	90.1%	estimated ± 0.9 pp, medium confidence
3	GPT-5.6 Sol	90.0%	estimated ± 1.8 pp, low confidence
4	Qwen3.8-Omni-Flash	89.7%	estimated ± 0.9 pp, medium confidence
5	Kimi K3	89.7%	estimated ± 0.9 pp, medium confidence
6	Claude Opus 4.7 (Adaptive)	89.6%	estimated ± 0.9 pp, medium confidence
7	Qwen3.8-Flash-Next	89.5%	estimated ± 0.9 pp, medium confidence
8	Qwen3.8-27B	89.5%	estimated ± 0.9 pp, medium confidence
9	Claude Opus 4.8	89.4%	estimated ± 0.9 pp, medium confidence
10	GLM-5.3-Flash	89.3%	estimated ± 0.9 pp, medium confidence
11	Gemini 3.7 Flash	89.2%	estimated ± 0.9 pp, medium confidence
12	Muse Spark 1.1	89.1%	estimated ± 0.9 pp, medium confidence
13	GPT-5.5	89.1%	estimated ± 1.8 pp, medium confidence
14	Claude Sonnet 5	89.1%	estimated ± 0.9 pp, medium confidence
15	GPT-5.6 Terra	88.8%	estimated ± 1.8 pp, medium confidence
16	dots3-note Preview	88.8%	estimated ± 1.2 pp, medium confidence
17	Sakana Fugu-Ultra	88.7%	estimated ± 0.9 pp, medium confidence
18	Muse Spark	88.7%	estimated ± 0.9 pp, medium confidence
19	Kimi K2.5	88.6%	estimated ± 1.2 pp, medium confidence
20	Seed 2.1 Pro	88.5%	estimated ± 0.9 pp, medium confidence
21	Sakana Fugu	88.4%	estimated ± 0.9 pp, medium confidence
22	Gemini 3.5 Flash	88.3%	estimated ± 0.9 pp, medium confidence
23	Qwen3.7 Plus	88.0%	measured
24	GPT-5.4	88.0%	estimated ± 0.9 pp, medium confidence
25	Seed 2.1 Turbo	87.9%	estimated ± 0.9 pp, medium confidence
26	GPT-5.2	87.8%	estimated ± 0.9 pp, medium confidence
27	Inkling	87.8%	estimated ± 0.9 pp, medium confidence
28	Kimi K2.5 (Reasoning)	87.8%	estimated ± 1.8 pp, medium confidence
29	GPT-5.6 Luna	87.7%	estimated ± 1.8 pp, medium confidence
30	Qwen3.6 Plus	87.7%	estimated ± 0.9 pp, medium confidence
31	MiMo-V2.5	87.7%	measured
32	Qwen3.6-27B	87.7%	measured
33	Gemini 3 Pro	87.7%	estimated ± 0.9 pp, medium confidence
34	Inkling-Small	87.7%	estimated ± 0.9 pp, medium confidence
35	Grok 4.3	87.6%	estimated ± 1.8 pp, medium confidence
36	Qwen3.5 397B	87.6%	estimated ± 0.9 pp, medium confidence
37	Pareto 26.9	87.6%	estimated ± 1.8 pp, medium confidence
38	Kimi K2.6	87.5%	estimated ± 0.9 pp, medium confidence
39	Gemini 3.1 Pro	87.4%	estimated ± 0.9 pp, medium confidence
40	Claude Opus 4.6	87.2%	estimated ± 1.8 pp, medium confidence
41	Muse Glimmer 30B	87.2%	estimated ± 0.9 pp, medium confidence
42	Gemma 4 31B	87.1%	estimated ± 1.8 pp, medium confidence
43	GPT-5.4 mini	86.9%	estimated ± 1.8 pp, medium confidence
44	Claude Sonnet 4.6	86.9%	estimated ± 0.9 pp, low confidence
45	Qwen3.5-122B-A10B	86.8%	estimated ± 0.9 pp, low confidence
46	Step 5 Preview	86.7%	estimated ± 1.8 pp, medium confidence
47	Nemotron 3 Nano Omni 30B A3B	86.6%	estimated ± 0.9 pp, low confidence
48	Qwen3.6-35B-A3B	86.6%	measured
49	Gemini 3.1 Flash-Lite	86.0%	estimated ± 0.9 pp, low confidence
50	Gemma 4 26B A4B	85.8%	estimated ± 1.8 pp, low confidence
51	MiniMax M3	85.4%	measured
52	Claude Opus 4.5	85.1%	estimated ± 0.9 pp, low confidence
53	Interfaze Beta	84.8%	estimated ± 1.8 pp, low confidence
54	Gemma 4 12B	84.2%	estimated ± 1.8 pp, low confidence
55	Grok 4.20	83.5%	estimated ± 0.9 pp, low confidence
56	GPT-5.4 nano	83.4%	estimated ± 1.8 pp, low confidence
57	Command A+	81.9%	estimated ± 0.9 pp, low confidence
58	LFM2.5-VL-3B	80.4%	estimated ± 1.8 pp, low confidence

Video-MME (with subtitle) leaderboard

All leaderboards

Agentic · terminal

Agentic · tools

Coding

Math

Knowledge & reasoning

Instruction following

Multilingual

Vision & documents

Long context

General