benchgap
Vision & documents

V* leaderboard

As of 2026-11-21, the highest measured score on V* is 96.9% by Kimi K2.6. 89 more models have estimated scores, calibrated from the benchmarks they were measured on.

Measured scores: benchlm.ai.

#ModelScoreSource
1Apodex 1.1100.0%estimated ± 7.5 pp, medium confidence
2Apodex 1.1 Mini100.0%estimated ± 7.5 pp, medium confidence
3Claude Opus 4.5 Thinking100.0%estimated ± 7.5 pp, medium confidence
4Claude Opus 4.6 (Adaptive)100.0%estimated ± 7.5 pp, medium confidence
5Claude Opus 4.7100.0%estimated ± 7.5 pp, medium confidence
6Claude Opus 5100.0%estimated ± 7.5 pp, low confidence
7Claude Opus 5.5100.0%estimated ± 7.5 pp, low confidence
8Claude Sonnet 5100.0%estimated ± 7.5 pp, medium confidence
9DeepSeek V4.1 Flash 1017100.0%estimated ± 7.5 pp, medium confidence
10Gemini 2.5 Pro100.0%estimated ± 7.5 pp, medium confidence
11Gemini 3.5 Flash100.0%estimated ± 7.5 pp, low confidence
12Gemini 3.5 Flash-Lite100.0%estimated ± 7.5 pp, medium confidence
13Gemini 3.6 Flash100.0%estimated ± 7.5 pp, low confidence
14Gemini 3.7 Flash100.0%estimated ± 7.5 pp, low confidence
15Gemini 3.8 Flash100.0%estimated ± 7.5 pp, low confidence
16Gemini 3 Flash100.0%estimated ± 7.5 pp, medium confidence
17Gemma 4 31B100.0%estimated ± 7.5 pp, medium confidence
18GPT-5100.0%estimated ± 7.5 pp, medium confidence
19GPT-5.1100.0%estimated ± 7.5 pp, medium confidence
20GPT-5.2-Codex100.0%estimated ± 7.5 pp, medium confidence
21GPT-5.3 Codex100.0%estimated ± 7.5 pp, medium confidence
22GPT-5.4 mini100.0%estimated ± 7.5 pp, medium confidence
23GPT-5.5100.0%estimated ± 7.5 pp, medium confidence
24GPT-5.6 Luna100.0%estimated ± 7.5 pp, medium confidence
25GPT-5.6 Sol100.0%estimated ± 7.5 pp, low confidence
26GPT-5.6 Terra100.0%estimated ± 7.5 pp, low confidence
27GPT-6.1 Sol100.0%estimated ± 7.5 pp, low confidence
28GPT-6.1 Terra100.0%estimated ± 7.5 pp, low confidence
29GPT-6 Luna100.0%estimated ± 7.5 pp, medium confidence
30GPT-6 Sol100.0%estimated ± 7.5 pp, low confidence
31Grok 4.3100.0%estimated ± 7.5 pp, medium confidence
32Grok 4.5100.0%estimated ± 7.5 pp, low confidence
33Inkling100.0%estimated ± 7.5 pp, medium confidence
34Inkling-Small100.0%estimated ± 7.5 pp, medium confidence
35Kimi K3.1 Pro100.0%estimated ± 7.5 pp, low confidence
36Kimi K3100.0%estimated ± 7.5 pp, low confidence
37Ling 3.0 Flash VL100.0%estimated ± 7.5 pp, medium confidence
38MiMo-V2.6-Flash100.0%estimated ± 7.5 pp, medium confidence
39MiniMax M3100.0%estimated ± 7.5 pp, medium confidence
40MiniMax M3.1 Flash100.0%estimated ± 7.5 pp, medium confidence
41Muse Spark Open100.0%estimated ± 7.5 pp, low confidence
42Qwen3.6-35B-A3B100.0%estimated ± 7.5 pp, medium confidence
43Qwen3.8-27B100.0%estimated ± 7.5 pp, medium confidence
44Qwen3.8-Flash-Next100.0%estimated ± 7.5 pp, medium confidence
45Qwen3.8 Max Preview100.0%estimated ± 7.5 pp, low confidence
46Step 5 Preview100.0%estimated ± 7.5 pp, medium confidence
47Glimpse99.8%estimated ± 11.2 pp, low confidence
48Kimi K2.599.0%estimated ± 5.8 pp, low confidence
49Kimi K2.696.9%measured
50Qwen3.6 Plus96.9%measured
51GPT-6 Astra96.5%estimated ± 6.5 pp, low confidence
52Claude Opus 4.896.3%estimated ± 6.5 pp, low confidence
53GPT-5.496.2%estimated ± 6.5 pp, low confidence
54Qwen3.8 Max96.1%estimated ± 6.5 pp, low confidence
55Gemini 3.1 Pro96.1%estimated ± 6.5 pp, low confidence
56Muse Spark96.1%estimated ± 6.5 pp, low confidence
57Inkling 2 Small96.0%estimated ± 6.5 pp, low confidence
58Claude Opus 4.696.0%estimated ± 6.5 pp, low confidence
59Qwen3.5 397B95.8%measured
60Qwen3.7 Plus95.6%estimated ± 6.5 pp, low confidence
61Step 3.7 Flash95.3%measured
62Inkling 2 Mini95.3%estimated ± 6.5 pp, low confidence
63Muse Glimmer 30B95.1%estimated ± 6.5 pp, low confidence
64Qwen3.6-27B94.7%measured
65GLM-5V-Turbo94.6%estimated ± 7.5 pp, medium confidence
66Qwen3.8-Omni-Flash94.2%estimated ± 10.1 pp, low confidence
67Holo2-235B-A22B94.1%estimated ± 6.5 pp, low confidence
68Qwen3.5-27B93.7%measured
69Seed 2.1 Pro93.4%estimated ± 11.2 pp, low confidence
70Qwen3.5-122B-A10B93.2%measured
71Seed 2.1 Turbo93.0%estimated ± 11.2 pp, low confidence
72Holo2-30B-A3B92.8%estimated ± 6.5 pp, low confidence
73Qwen3.5-35B-A3B92.7%measured
74Qwen3-VL-235B-A22B Instruct90.9%estimated ± 6.5 pp, low confidence
75GPT-5.1-Codex89.0%estimated ± 7.5 pp, medium confidence
76GPT-5.1-Codex-Max89.0%estimated ± 7.5 pp, medium confidence
77Holo2-8B88.8%estimated ± 6.5 pp, low confidence
78Gemini 3 Pro88.0%measured
79Nemotron 3 Nano Omni 30B A3B87.9%estimated ± 6.5 pp, low confidence
80Holo2-4B87.3%estimated ± 6.5 pp, low confidence
81LFM3-VL-2B86.4%measured
82Nemotron 3.5 Super VL 120B A12B NVFP486.0%measured
83Nemotron 3.5 Super VL 120B A12B85.6%measured
84Ternary Bonsai 2 27B81.6%estimated ± 10.1 pp, low confidence
85Grok 4.2081.3%estimated ± 11.2 pp, low confidence
86GPT-5.275.9%measured
87Claude Opus 4.567.0%measured
88Claude Sonnet 4.658.0%estimated ± 7.5 pp, low confidence
89ZAYA1-VL-8B53.5%estimated ± 5.8 pp, low confidence
90o351.0%estimated ± 7.5 pp, low confidence
91LFM2.5-VL-3B50.2%estimated ± 5.8 pp, low confidence
92MiMo-V2-Omni48.3%estimated ± 7.5 pp, low confidence
93Gemma 4 12B45.6%estimated ± 7.5 pp, low confidence
94Gemma 4 26B A4B39.3%estimated ± 7.5 pp, low confidence
95o4-mini39.3%estimated ± 7.5 pp, low confidence
96Grok 434.5%estimated ± 7.5 pp, low confidence
97Gemini 2.5 Flash1.3%estimated ± 7.5 pp, low confidence
98GPT-5.4 nano0.5%estimated ± 7.5 pp, low confidence
99Qwen3.5-4B0.5%estimated ± 7.5 pp, low confidence
100Claude 4 Sonnet0.0%estimated ± 7.5 pp, low confidence
101Command A+0.0%estimated ± 7.5 pp, low confidence
102Grok 4.1 Fast (Reasoning)0.0%estimated ± 7.5 pp, low confidence
103Mistral Medium 3.5 128B0.0%estimated ± 7.5 pp, low confidence

All leaderboards

Agentic · terminal

Agentic · tools

Coding

Math

Knowledge & reasoning

Instruction following

Multilingual

Vision & documents

Long context

General