Vision & documents

Video-MME (w/o subtitle) leaderboard

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

As of 2026-10-10, the highest measured score on Video-MME (w/o subtitle) is 87.7% by Gemini 3 Pro. 111 more models have estimated scores, calibrated from the benchmarks they were measured on.

Measured scores: benchlm.ai.

#	Model	Score	Source
1	Step 3.7 Flash	100.0%	estimated ± 3.2 pp, low confidence
2	Claude Opus 5.5	90.8%	estimated ± 6.3 pp, low confidence
3	GPT-6 Astra	90.2%	estimated ± 6.3 pp, low confidence
4	Qwen3.8 Max	89.7%	estimated ± 1.7 pp, low confidence
5	GPT-6.1 Sol	89.7%	estimated ± 6.3 pp, low confidence
6	Gemini 3.8 Flash	89.4%	estimated ± 6.3 pp, low confidence
7	Gemini 3.7 Flash	89.3%	estimated ± 6.3 pp, low confidence
8	Claude Opus 5	88.8%	estimated ± 6.3 pp, low confidence
9	Gemini 3.6 Flash	87.9%	estimated ± 6.3 pp, low confidence
10	GPT-6 Sol	87.8%	estimated ± 6.3 pp, low confidence
11	Gemini 3 Pro	87.7%	measured
12	Qwen3.8 Max Preview	87.7%	estimated ± 6.3 pp, low confidence
13	Gemini 3.1 Pro	87.3%	estimated ± 1.8 pp, low confidence
14	Gemini 3.5 Flash	87.1%	estimated ± 1.8 pp, low confidence
15	GPT-5.6 Sol	86.8%	estimated ± 1.8 pp, low confidence
16	Qwen3.8-Omni-Flash	86.5%	estimated ± 3.0 pp, low confidence
17	Grok 4.5	86.3%	estimated ± 6.3 pp, low confidence
18	dots3-note Preview	86.3%	estimated ± 1.7 pp, medium confidence
19	Qwen3.8-Flash-Next	86.2%	estimated ± 3.0 pp, low confidence
20	Kimi K3	86.2%	estimated ± 1.8 pp, low confidence
21	Seed 2.1 Pro	86.2%	estimated ± 1.8 pp, low confidence
22	Qwen3.8-27B	86.1%	estimated ± 3.0 pp, low confidence
23	GPT-5.4	86.0%	estimated ± 1.8 pp, low confidence
24	GPT-5.5	86.0%	estimated ± 1.8 pp, low confidence
25	GPT-6 Luna	85.9%	estimated ± 6.3 pp, low confidence
26	GPT-5.2	85.8%	measured
27	GPT-5.6 Terra	85.7%	estimated ± 1.8 pp, medium confidence
28	Apodex 1.1	85.6%	estimated ± 6.3 pp, low confidence
29	Apodex 1.1 Mini	85.6%	estimated ± 6.3 pp, low confidence
30	Muse Spark	85.6%	estimated ± 1.8 pp, medium confidence
31	Gemini 3.5 Flash-Lite	85.5%	estimated ± 6.3 pp, low confidence
32	Ling 3.0 Flash VL	85.5%	estimated ± 6.3 pp, low confidence
33	Qwen3.6 Plus	85.5%	estimated ± 1.0 pp, medium confidence
34	Seed 2.1 Turbo	85.4%	estimated ± 1.8 pp, medium confidence
35	Claude Opus 4.7 (Adaptive)	85.4%	estimated ± 6.3 pp, low confidence
36	Gemini 3 Flash	85.3%	estimated ± 6.3 pp, low confidence
37	GPT-5.3 Codex	85.3%	estimated ± 6.3 pp, low confidence
38	Kimi K2.6	85.1%	estimated ± 1.8 pp, medium confidence
39	Kimi K2.5 (Reasoning)	84.7%	estimated ± 1.8 pp, medium confidence
40	Claude Sonnet 5	84.7%	estimated ± 6.3 pp, low confidence
41	GPT-5.6 Luna	84.6%	estimated ± 1.8 pp, medium confidence
42	DeepSeek V4.1 Flash	84.5%	estimated ± 6.3 pp, low confidence
43	Grok 4.3	84.5%	estimated ± 1.8 pp, medium confidence
44	Pareto 26.9	84.4%	estimated ± 1.8 pp, medium confidence
45	MiMo-V2.5	84.4%	estimated ± 1.8 pp, medium confidence
46	Claude Opus 4.7	84.2%	estimated ± 6.3 pp, low confidence
47	Mistral Large 4	84.2%	estimated ± 6.3 pp, low confidence
48	GPT-5.2-Codex	84.2%	estimated ± 6.3 pp, low confidence
49	Claude Opus 4.6	84.1%	estimated ± 1.8 pp, medium confidence
50	Gemma 4 31B	83.9%	estimated ± 1.8 pp, medium confidence
51	Qwen3.7 Plus	83.9%	estimated ± 1.7 pp, medium confidence
52	GPT-5.1	83.8%	estimated ± 6.3 pp, low confidence
53	Claude Opus 4.6 (Adaptive)	83.7%	estimated ± 6.3 pp, low confidence
54	GPT-5.4 mini	83.7%	estimated ± 1.8 pp, medium confidence
55	Qwen3.5-122B-A10B	83.7%	estimated ± 1.0 pp, medium confidence
56	Qwen3.5 397B	83.7%	measured
57	Gemini 2.5 Pro	83.5%	estimated ± 6.3 pp, low confidence
58	Step 5 Preview	83.4%	estimated ± 1.8 pp, medium confidence
59	GPT-5 (medium)	83.2%	estimated ± 6.3 pp, low confidence
60	Kimi K2.5	83.2%	measured
61	GPT-5 (high)	83.2%	estimated ± 6.3 pp, low confidence
62	Claude Opus 4.5 Thinking	83.1%	estimated ± 6.3 pp, low confidence
63	Grok 4.20	83.0%	estimated ± 1.8 pp, medium confidence
64	Qwen3.6-27B	82.8%	estimated ± 1.0 pp, medium confidence
65	MiMo-V2.6-Flash	82.7%	estimated ± 6.3 pp, low confidence
66	GLM-5V-Turbo	82.6%	estimated ± 6.3 pp, low confidence
67	MiniMax M3	82.5%	estimated ± 1.7 pp, medium confidence
68	Qwen3.6-35B-A3B	82.5%	measured
69	GPT-5.1-Codex	82.4%	estimated ± 6.3 pp, low confidence
70	GPT-5.1-Codex-Max	82.4%	estimated ± 6.3 pp, low confidence
71	Inkling-Small	82.4%	estimated ± 1.8 pp, medium confidence
72	Muse Glimmer 30B	82.4%	estimated ± 1.8 pp, medium confidence
73	Qwen3.5-27B	82.3%	estimated ± 1.0 pp, medium confidence
74	Gemma 4 26B A4B	82.3%	estimated ± 1.8 pp, medium confidence
75	Inkling	82.1%	estimated ± 1.8 pp, medium confidence
76	Claude Sonnet 4.6	81.7%	estimated ± 6.3 pp, low confidence
77	Qwen3.5-35B-A3B	81.5%	estimated ± 1.0 pp, medium confidence
78	o3	81.5%	estimated ± 6.3 pp, low confidence
79	MiMo-V2-Omni	81.4%	estimated ± 6.3 pp, low confidence
80	Claude Opus 4.5	81.4%	measured
81	Grok 4	81.0%	estimated ± 6.3 pp, low confidence
82	Interfaze Beta	80.9%	estimated ± 1.8 pp, medium confidence
83	Claude 4.1 Opus Thinking	80.7%	estimated ± 6.3 pp, low confidence
84	Gemini 2.5 Flash	80.0%	estimated ± 6.3 pp, low confidence
85	Gemma 4 12B	79.9%	estimated ± 1.8 pp, low confidence
86	Mistral Medium 3.5 128B	79.8%	estimated ± 6.3 pp, low confidence
87	Grok 4.1 Fast (Reasoning)	79.4%	estimated ± 6.3 pp, low confidence
88	Claude 4 Sonnet	79.2%	estimated ± 6.3 pp, low confidence
89	Llama 4 Maverick	79.1%	estimated ± 6.3 pp, low confidence
90	Grok 4 Fast (Reasoning)	79.1%	estimated ± 6.3 pp, low confidence
91	GPT-4.1	78.9%	estimated ± 6.3 pp, low confidence
92	Qwen3-Omni-30B-A3B-Thinking	78.7%	estimated ± 6.3 pp, low confidence
93	GPT-4.1 mini	78.5%	estimated ± 6.3 pp, low confidence
94	GPT-5.4 nano	78.3%	estimated ± 1.8 pp, low confidence
95	Mistral Small 4	78.2%	estimated ± 6.3 pp, low confidence
96	Mistral Small 4 (Reasoning)	78.2%	estimated ± 6.3 pp, low confidence
97	Mistral Large 3	78.0%	estimated ± 6.3 pp, low confidence
98	Qwen3-Omni-30B-A3B-Instruct	78.0%	estimated ± 6.3 pp, low confidence
99	Gemini 1.5 Pro	77.9%	estimated ± 6.3 pp, low confidence
100	Mistral Medium 3	77.7%	estimated ± 6.3 pp, low confidence
101	Llama 4 Scout	77.6%	estimated ± 6.3 pp, low confidence
102	Qwen3.5 397B (Reasoning)	77.6%	estimated ± 6.3 pp, low confidence
103	Grok 4.1 Fast	77.2%	estimated ± 6.3 pp, low confidence
104	Gemma 3 27B	77.2%	estimated ± 6.3 pp, low confidence
105	Nova Pro	77.0%	estimated ± 6.3 pp, low confidence
106	GPT-4o mini	76.9%	estimated ± 6.3 pp, low confidence
107	GPT-4.1 nano	76.9%	estimated ± 6.3 pp, low confidence
108	Claude 3 Haiku	76.7%	estimated ± 6.3 pp, low confidence
109	LFM2.5-VL-1.6B-Extract	76.7%	estimated ± 6.3 pp, low confidence
110	Phi-4 Multimodal Instruct	76.7%	estimated ± 6.3 pp, low confidence
111	Command A+	76.1%	estimated ± 1.0 pp, medium confidence
112	Nemotron 3 Nano Omni 30B A3B	72.2%	measured
113	Gemma 4 E4B	70.7%	estimated ± 1.8 pp, low confidence
114	Qwen2.5-VL-32B	68.8%	estimated ± 1.8 pp, low confidence
115	Gemma 4 E2B	65.5%	estimated ± 1.8 pp, low confidence
116	LFM2.5-VL-3B	52.9%	estimated ± 1.0 pp, low confidence
117	ZAYA1-VL-8B	50.8%	estimated ± 1.0 pp, low confidence
118	LFM2.5-VL-450M	39.2%	estimated ± 1.0 pp, low confidence

Video-MME (w/o subtitle) leaderboard

All leaderboards

Agentic · terminal

Agentic · tools

Coding

Math

Knowledge & reasoning

Instruction following

Multilingual

Vision & documents

Long context

General