Knowledge & reasoning
HLE-Verified leaderboard
As of 2026-10-07, the highest measured score on HLE-Verified is 54.9% by Gemini 3.8 Flash. 206 more models have estimated scores, calibrated from the benchmarks they were measured on.
Measured scores: benchlm.ai.
| # | Model | Score | Source |
|---|---|---|---|
| 1 | Gemini 3.8 Flash | 54.9% | measured |
| 2 | Gemini 4 Argon | 54.8% | estimated ± 1.2 pp, low confidence |
| 3 | Claude Fable 5.1 | 54.6% | estimated ± 0.7 pp, medium confidence |
| 4 | Claude Opus 5.5 | 54.6% | estimated ± 0.7 pp, medium confidence |
| 5 | GPT-6.1 Sol | 54.6% | estimated ± 0.7 pp, low confidence |
| 6 | GPT-6 Astra | 54.6% | estimated ± 0.7 pp, low confidence |
| 7 | GPT-6 Sol | 54.6% | estimated ± 0.7 pp, medium confidence |
| 8 | Claude Fable 5 | 54.6% | estimated ± 0.7 pp, medium confidence |
| 9 | GPT-5.6 Sol | 54.5% | measured |
| 10 | Claude Opus 5 | 54.4% | measured |
| 11 | Qwen3.8 Max | 54.3% | estimated ± 4.6 pp, medium confidence |
| 12 | GPT-5.5 | 54.1% | estimated ± 0.7 pp, medium confidence |
| 13 | Muse Spark 1.3 | 54.1% | estimated ± 2.0 pp, medium confidence |
| 14 | Claude Haiku 5.5 | 54.0% | estimated ± 6.8 pp, low confidence |
| 15 | Gemini 3.7 Flash | 53.6% | measured |
| 16 | MiniMax M3 | 53.0% | estimated ± 2.0 pp, medium confidence |
| 17 | Qwen3.8 Max Preview | 52.7% | estimated ± 2.0 pp, medium confidence |
| 18 | GPT-5.5 Pro | 52.5% | estimated ± 0.7 pp, medium confidence |
| 19 | GPT-5.6 Terra | 51.1% | measured |
| 20 | Claude Sonnet 5.5 | 50.6% | estimated ± 6.0 pp, low confidence |
| 21 | Qwen3.7 Max | 49.9% | estimated ± 2.0 pp, medium confidence |
| 22 | Qwen3.8-Flash-Next | 49.9% | estimated ± 2.0 pp, medium confidence |
| 23 | GPT-5.4 Pro | 46.4% | estimated ± 0.7 pp, low confidence |
| 24 | Grok 4.7 | 42.5% | estimated ± 6.0 pp, low confidence |
| 25 | GLM-5.3 | 42.3% | estimated ± 2.0 pp, medium confidence |
| 26 | DeepSeek V4.1 Flash | 41.3% | estimated ± 6.0 pp, low confidence |
| 27 | GPT-5.3 Codex | 38.7% | estimated ± 2.0 pp, medium confidence |
| 28 | dots3-note Preview | 35.9% | estimated ± 0.7 pp, low confidence |
| 29 | Step 5 Preview | 35.8% | estimated ± 6.0 pp, low confidence |
| 30 | Gemini 3.1 Pro | 35.3% | estimated ± 0.7 pp, low confidence |
| 31 | Claude Opus 4.5 Thinking | 35.3% | estimated ± 0.7 pp, low confidence |
| 32 | Claude Opus 4.6 (Adaptive) | 35.3% | estimated ± 0.7 pp, low confidence |
| 33 | Claude Opus 4.7 (Adaptive) | 35.3% | estimated ± 0.7 pp, low confidence |
| 34 | Claude Opus 4.8 | 35.3% | estimated ± 0.7 pp, low confidence |
| 35 | Claude Sonnet 4.5 | 35.3% | estimated ± 0.7 pp, low confidence |
| 36 | Claude Sonnet 4.5 Thinking | 35.3% | estimated ± 0.7 pp, low confidence |
| 37 | Claude Sonnet 4.6 | 35.3% | estimated ± 0.7 pp, low confidence |
| 38 | DeepSeek V4 Flash 0731 | 35.3% | estimated ± 0.7 pp, low confidence |
| 39 | DeepSeek V4 Pro 0813 | 35.3% | estimated ± 0.7 pp, low confidence |
| 40 | Gemini 3.5 Flash | 35.3% | estimated ± 0.7 pp, low confidence |
| 41 | Gemini 3.6 Flash | 35.3% | estimated ± 0.7 pp, low confidence |
| 42 | Gemini 3 Pro | 35.3% | estimated ± 0.7 pp, low confidence |
| 43 | Gemini 3 Pro Deep Think | 35.3% | estimated ± 0.7 pp, low confidence |
| 44 | GPT-5.2 | 35.3% | estimated ± 0.7 pp, low confidence |
| 45 | GPT-5.4 | 35.3% | estimated ± 0.7 pp, low confidence |
| 46 | GPT-5.4 mini | 35.3% | estimated ± 0.7 pp, low confidence |
| 47 | GPT-5.4 nano | 35.3% | estimated ± 0.7 pp, low confidence |
| 48 | GPT-5.6 Luna | 35.3% | estimated ± 0.7 pp, low confidence |
| 49 | GPT-6 Luna | 35.3% | estimated ± 0.7 pp, low confidence |
| 50 | Grok 4.20 | 35.3% | estimated ± 0.7 pp, low confidence |
| 51 | Grok 4.5 | 35.3% | estimated ± 0.7 pp, low confidence |
| 52 | Grok 4.6 | 35.3% | estimated ± 0.7 pp, low confidence |
| 53 | Inkling-Small | 35.3% | estimated ± 0.7 pp, low confidence |
| 54 | Kimi K3 | 35.3% | estimated ± 0.7 pp, low confidence |
| 55 | Muse Spark | 35.3% | estimated ± 0.7 pp, low confidence |
| 56 | GLM-5.3-Flash | 32.9% | estimated ± 2.0 pp, medium confidence |
| 57 | Claude Sonnet 5 | 31.0% | measured |
| 58 | Kimi K2.6 | 31.0% | estimated ± 2.0 pp, medium confidence |
| 59 | MiMo-V2.6-Pro | 28.8% | estimated ± 6.0 pp, low confidence |
| 60 | Ling 3.1 Flash | 23.3% | estimated ± 6.0 pp, low confidence |
| 61 | Qwen3.8-27B | 21.9% | estimated ± 2.0 pp, low confidence |
| 62 | MiMo-V2.6-Flash | 21.4% | estimated ± 6.0 pp, low confidence |
| 63 | Muse Spark 1.2 | 20.8% | estimated ± 2.0 pp, low confidence |
| 64 | Mistral Large 4 | 20.3% | estimated ± 6.0 pp, low confidence |
| 65 | Grok 4.3 | 18.6% | estimated ± 2.0 pp, low confidence |
| 66 | Qwen3.7 Plus | 18.0% | estimated ± 2.0 pp, low confidence |
| 67 | GPT-5.2-Codex | 17.6% | estimated ± 2.0 pp, low confidence |
| 68 | Muse Spark 1.1 | 17.2% | estimated ± 2.0 pp, low confidence |
| 69 | Mercury 2.5 | 16.9% | estimated ± 6.0 pp, low confidence |
| 70 | Hy3 | 16.8% | estimated ± 2.0 pp, low confidence |
| 71 | Hy3 Preview | 16.8% | estimated ± 2.0 pp, low confidence |
| 72 | Kimi K2.7 Code | 16.6% | estimated ± 2.0 pp, low confidence |
| 73 | GLM-5.2 | 16.4% | estimated ± 2.0 pp, low confidence |
| 74 | Solar Pro 4 | 15.8% | estimated ± 2.0 pp, low confidence |
| 75 | Qwen 3.6 Max (preview) | 15.6% | estimated ± 2.0 pp, low confidence |
| 76 | Claude Opus 4.7 | 15.5% | estimated ± 2.0 pp, low confidence |
| 77 | Qwen3.6 Plus | 15.4% | estimated ± 2.0 pp, low confidence |
| 78 | Kimi K2.5 | 15.4% | estimated ± 2.0 pp, low confidence |
| 79 | Kimi K2.5 (Reasoning) | 15.4% | estimated ± 2.0 pp, low confidence |
| 80 | Grok 4 | 15.3% | estimated ± 2.0 pp, low confidence |
| 81 | MiniMax M2.7 | 15.3% | estimated ± 2.0 pp, low confidence |
| 82 | GPT-5.1 | 15.3% | estimated ± 2.0 pp, low confidence |
| 83 | Inkling | 15.3% | estimated ± 2.0 pp, low confidence |
| 84 | MiMo-V2-Pro | 15.3% | estimated ± 2.0 pp, low confidence |
| 85 | GLM-5.1 | 15.3% | estimated ± 2.0 pp, low confidence |
| 86 | Nemotron 3 Ultra | 15.3% | estimated ± 2.0 pp, low confidence |
| 87 | MiMo-V2.5-Pro | 15.3% | estimated ± 2.0 pp, low confidence |
| 88 | Apodex 1.1 | 15.3% | estimated ± 2.0 pp, low confidence |
| 89 | Apodex 1.1 Mini | 15.3% | estimated ± 2.0 pp, low confidence |
| 90 | Ling 3.0 Flash VL | 15.3% | estimated ± 2.0 pp, low confidence |
| 91 | Qwen3.5 397B | 15.3% | estimated ± 2.0 pp, low confidence |
| 92 | Qwen3.5 397B (Reasoning) | 15.3% | estimated ± 2.0 pp, low confidence |
| 93 | GPT-5.1-Codex | 15.3% | estimated ± 2.0 pp, low confidence |
| 94 | GPT-5.1-Codex-Max | 15.3% | estimated ± 2.0 pp, low confidence |
| 95 | GLM-4.7 | 15.3% | estimated ± 2.0 pp, low confidence |
| 96 | Qwen3.5-27B | 15.3% | estimated ± 2.0 pp, low confidence |
| 97 | A.X K2 | 15.3% | estimated ± 2.0 pp, low confidence |
| 98 | Gemma 4 31B | 15.3% | estimated ± 2.0 pp, low confidence |
| 99 | Qwen3.5-122B-A10B | 15.3% | estimated ± 2.0 pp, low confidence |
| 100 | GPT-5 (high) | 15.3% | estimated ± 2.0 pp, low confidence |
| 101 | Ling 3.0 Flash | 15.3% | estimated ± 2.0 pp, low confidence |
| 102 | Ling 3.0 Flash FP8 | 15.3% | estimated ± 2.0 pp, low confidence |
| 103 | Grok 4.1 Fast (Reasoning) | 15.3% | estimated ± 2.0 pp, low confidence |
| 104 | Celeris-1 | 15.3% | estimated ± 2.0 pp, low confidence |
| 105 | Claude 3 Haiku | 15.3% | estimated ± 2.0 pp, low confidence |
| 106 | Claude 3 Opus | 15.3% | estimated ± 2.0 pp, low confidence |
| 107 | Claude 4.1 Opus Thinking | 15.3% | estimated ± 2.0 pp, low confidence |
| 108 | Claude 4 Sonnet | 15.3% | estimated ± 2.0 pp, low confidence |
| 109 | Claude Opus 4.5 | 15.3% | estimated ± 2.0 pp, low confidence |
| 110 | Claude Opus 4.6 | 15.3% | estimated ± 2.0 pp, low confidence |
| 111 | Command A+ | 15.3% | estimated ± 2.0 pp, low confidence |
| 112 | DeepSeek-R1 | 15.3% | estimated ± 2.0 pp, low confidence |
| 113 | DeepSeek R1 Distill Qwen 32B | 15.3% | estimated ± 2.0 pp, low confidence |
| 114 | DeepSeek V3 | 15.3% | estimated ± 2.0 pp, low confidence |
| 115 | DeepSeek V3 0324 | 15.3% | estimated ± 2.0 pp, low confidence |
| 116 | DeepSeek V3.1 | 15.3% | estimated ± 2.0 pp, low confidence |
| 117 | DeepSeek V3.1 (Reasoning) | 15.3% | estimated ± 2.0 pp, low confidence |
| 118 | DeepSeek V3.2 | 15.3% | estimated ± 2.0 pp, low confidence |
| 119 | Exaone 4.0 1.2B | 15.3% | estimated ± 2.0 pp, low confidence |
| 120 | Exaone 4.0 32B | 15.3% | estimated ± 2.0 pp, low confidence |
| 121 | Gemini 1.0 Pro | 15.3% | estimated ± 2.0 pp, low confidence |
| 122 | Gemini 1.5 Pro | 15.3% | estimated ± 2.0 pp, low confidence |
| 123 | Gemini 2.5 Flash | 15.3% | estimated ± 2.0 pp, low confidence |
| 124 | Gemini 2.5 Pro | 15.3% | estimated ± 2.0 pp, low confidence |
| 125 | Gemini 3.5 Flash-Lite | 15.3% | estimated ± 2.0 pp, low confidence |
| 126 | Gemini 3 Flash | 15.3% | estimated ± 2.0 pp, low confidence |
| 127 | Gemma 3 27B | 15.3% | estimated ± 2.0 pp, low confidence |
| 128 | Gemma 4 12B | 15.3% | estimated ± 2.0 pp, low confidence |
| 129 | Gemma 4 26B A4B | 15.3% | estimated ± 2.0 pp, low confidence |
| 130 | Gemma 4 E2B | 15.3% | estimated ± 2.0 pp, low confidence |
| 131 | Gemma 4 E4B | 15.3% | estimated ± 2.0 pp, low confidence |
| 132 | GLM-4.5-Air | 15.3% | estimated ± 2.0 pp, low confidence |
| 133 | GLM-4.6 | 15.3% | estimated ± 2.0 pp, low confidence |
| 134 | GLM-5 | 15.3% | estimated ± 2.0 pp, low confidence |
| 135 | GLM-5-Turbo | 15.3% | estimated ± 2.0 pp, low confidence |
| 136 | GLM-5V-Turbo | 15.3% | estimated ± 2.0 pp, low confidence |
| 137 | GPT-4.1 | 15.3% | estimated ± 2.0 pp, low confidence |
| 138 | GPT-4.1 mini | 15.3% | estimated ± 2.0 pp, low confidence |
| 139 | GPT-4.1 nano | 15.3% | estimated ± 2.0 pp, low confidence |
| 140 | GPT-4o | 15.3% | estimated ± 2.0 pp, low confidence |
| 141 | GPT-4o mini | 15.3% | estimated ± 2.0 pp, low confidence |
| 142 | GPT-5 (medium) | 15.3% | estimated ± 2.0 pp, low confidence |
| 143 | GPT-OSS 120B | 15.3% | estimated ± 2.0 pp, low confidence |
| 144 | GPT-OSS 20B | 15.3% | estimated ± 2.0 pp, low confidence |
| 145 | Granite-4.0-350M | 15.3% | estimated ± 2.0 pp, low confidence |
| 146 | Granite-4.0-H-1B | 15.3% | estimated ± 2.0 pp, low confidence |
| 147 | Granite-4.0-H-350M | 15.3% | estimated ± 2.0 pp, low confidence |
| 148 | Granite 4.2 30B | 15.3% | estimated ± 2.0 pp, low confidence |
| 149 | Granite 4.2 3B | 15.3% | estimated ± 2.0 pp, low confidence |
| 150 | Granite 4.2 8B | 15.3% | estimated ± 2.0 pp, low confidence |
| 151 | Grok 4.1 Fast | 15.3% | estimated ± 2.0 pp, low confidence |
| 152 | Grok 4 Fast (Reasoning) | 15.3% | estimated ± 2.0 pp, low confidence |
| 153 | Grok Code Fast 1 | 15.3% | estimated ± 2.0 pp, low confidence |
| 154 | K-Exaone | 15.3% | estimated ± 2.0 pp, low confidence |
| 155 | K-EXAONE 2.0 | 15.3% | estimated ± 2.0 pp, low confidence |
| 156 | Kimi K2 | 15.3% | estimated ± 2.0 pp, low confidence |
| 157 | LFM2.5-2.6B | 15.3% | estimated ± 2.0 pp, low confidence |
| 158 | LFM2.5-8B-A1B | 15.3% | estimated ± 2.0 pp, low confidence |
| 159 | LFM2.5-VL-1.6B-Extract | 15.3% | estimated ± 2.0 pp, low confidence |
| 160 | Ling 2.6 Flash | 15.3% | estimated ± 2.0 pp, low confidence |
| 161 | Ling 3.0 Tiny | 15.3% | estimated ± 2.0 pp, low confidence |
| 162 | Llama 3.1 405B | 15.3% | estimated ± 2.0 pp, low confidence |
| 163 | Llama 4 Maverick | 15.3% | estimated ± 2.0 pp, low confidence |
| 164 | Llama 4 Scout | 15.3% | estimated ± 2.0 pp, low confidence |
| 165 | MiMo-V2-Flash | 15.3% | estimated ± 2.0 pp, low confidence |
| 166 | MiMo-V2-Omni | 15.3% | estimated ± 2.0 pp, low confidence |
| 167 | MiniCPM5-2B | 15.3% | estimated ± 2.0 pp, low confidence |
| 168 | Mistral Large 2 | 15.3% | estimated ± 2.0 pp, low confidence |
| 169 | Mistral Large 3 | 15.3% | estimated ± 2.0 pp, low confidence |
| 170 | Mistral Medium 3 | 15.3% | estimated ± 2.0 pp, low confidence |
| 171 | Mistral Medium 3.5 128B | 15.3% | estimated ± 2.0 pp, low confidence |
| 172 | Mistral Small 4 | 15.3% | estimated ± 2.0 pp, low confidence |
| 173 | Mistral Small 4 (Reasoning) | 15.3% | estimated ± 2.0 pp, low confidence |
| 174 | Muse Glimmer 30B | 15.3% | estimated ± 2.0 pp, low confidence |
| 175 | Nemotron 3.5 Lightning 30B A3B NVFP4 | 15.3% | estimated ± 2.0 pp, low confidence |
| 176 | Nemotron 3 Nano 30B | 15.3% | estimated ± 2.0 pp, low confidence |
| 177 | Nemotron 3 Nano Omni 30B A3B | 15.3% | estimated ± 2.0 pp, low confidence |
| 178 | Nemotron 3 Super 100B | 15.3% | estimated ± 2.0 pp, low confidence |
| 179 | Nemotron Ultra 253B | 15.3% | estimated ± 2.0 pp, low confidence |
| 180 | North Mini Code | 15.3% | estimated ± 2.0 pp, low confidence |
| 181 | Nova Pro | 15.3% | estimated ± 2.0 pp, low confidence |
| 182 | o1 | 15.3% | estimated ± 2.0 pp, low confidence |
| 183 | o1-preview | 15.3% | estimated ± 2.0 pp, low confidence |
| 184 | o3 | 15.3% | estimated ± 2.0 pp, low confidence |
| 185 | o3-mini | 15.3% | estimated ± 2.0 pp, low confidence |
| 186 | o3-pro | 15.3% | estimated ± 2.0 pp, low confidence |
| 187 | Phi-4 | 15.3% | estimated ± 2.0 pp, low confidence |
| 188 | Phi-4 Multimodal Instruct | 15.3% | estimated ± 2.0 pp, low confidence |
| 189 | Quasar 438B | 15.3% | estimated ± 2.0 pp, low confidence |
| 190 | Qwen2.5 Coder 32B Instruct | 15.3% | estimated ± 2.0 pp, low confidence |
| 191 | Qwen3.5-35B-A3B | 15.3% | estimated ± 2.0 pp, low confidence |
| 192 | Qwen3.6-27B | 15.3% | estimated ± 2.0 pp, low confidence |
| 193 | Qwen3.6-35B-A3B | 15.3% | estimated ± 2.0 pp, low confidence |
| 194 | Qwen3 Max | 15.3% | estimated ± 2.0 pp, low confidence |
| 195 | Qwen3-Omni-30B-A3B-Instruct | 15.3% | estimated ± 2.0 pp, low confidence |
| 196 | Qwen3-Omni-30B-A3B-Thinking | 15.3% | estimated ± 2.0 pp, low confidence |
| 197 | Sarvam 105B | 15.3% | estimated ± 2.0 pp, low confidence |
| 198 | Sarvam 30B | 15.3% | estimated ± 2.0 pp, low confidence |
| 199 | Solar Pro 2 | 15.3% | estimated ± 2.0 pp, low confidence |
| 200 | Solar Pro 3 | 15.3% | estimated ± 2.0 pp, low confidence |
| 201 | Step 3.7 Flash | 15.3% | estimated ± 2.0 pp, low confidence |
| 202 | Trinity-Large-Preview | 15.3% | estimated ± 2.0 pp, low confidence |
| 203 | Trinity-Large-Thinking | 15.3% | estimated ± 2.0 pp, low confidence |
| 204 | Ultravox v0.6 Llama 3.3 70B | 15.3% | estimated ± 2.0 pp, low confidence |
| 205 | Qwen3.5 Flash | 0.0% | estimated ± 4.6 pp, low confidence |
| 206 | MiMo-V2.5 | 0.0% | estimated ± 4.6 pp, low confidence |
| 207 | Gemini 3.1 Flash-Lite | 0.0% | estimated ± 4.6 pp, low confidence |
| 208 | Claude Haiku 4.5 | 0.0% | estimated ± 4.6 pp, low confidence |
| 209 | GLM-4.5 | 0.0% | estimated ± 4.6 pp, low confidence |
| 210 | GPT-4 Turbo | 0.0% | estimated ± 6.8 pp, low confidence |
| 211 | Laguna M.1 | 0.0% | estimated ± 4.6 pp, low confidence |
| 212 | Laguna XS.2 | 0.0% | estimated ± 4.6 pp, low confidence |
All leaderboards
Agentic · terminal
- AA Terminal-Bench 2.1
- AA Terminal-Bench 4.0
- Terminal-Bench 2.0
- Terminal-Bench 3.0
- Terminal-Bench 4.0
- Terminal-Bench-Science 0.1
- Terminal-Bench 2.1
- Terminal-Bench 2.1 (Vals AI)
Agentic · tools
- AA Agentic Index
- AA-AnalystAgent
- AA AutomationBench
- AA EnterpriseOps-Gym
- GDP.pdf
- AA Harvey LAB
- AA ITBench
- AA Tau3 Banking
- Agents' Last Exam
- APEX-Agents
- APEX-Agents-AA
- ApprenticeBench
- AutomationBench
- BFCL v4
- BrowseComp
- Claw-Eval
- CWE-bench v1
- CyberGym
- DeepPlanning
- DeepSearchQA
- DRACO
- ExploitGym
- GDPval-AA
- Gert Labs
- HLE w/ tools
- JobBench
- MCP Atlas
- MCP-Tasks
- OSWorld-Verified
- OSWorld 2.0
- QwenClawBench
- ResearchClawBench
- SkillsBench
- τ²-bench results
- τ³-bench results
- Toolathlon
- Toolathlon-Verified
- VITA-Bench
- WideResearch
Coding
- AA Coding Index
- AA-SciCode
- CursorBench 3.1
- CursorBench 3.2
- CursorBench 4.0
- DeepSWE
- FrontierCode 1.1 Main
- FrontierCode 1.1 Extended
- FrontierSWE v2
- LiveCodeBench
- LiveCodeBench Pro
- LiveCodeBench v6
- NL2Repo
- OpenHarmony Bench
- PostTrainBench v1.1
- ProgramBench
- React Native Evals
- SciCode
- SWE-bench Pro
- SWE-bench Verified
- SWE-Rebench
- Vals LiveCodeBench
- Vals SWE-bench
- Vibe Code Bench
- VulcanBench v3
Math
- AIME 2025
- AIME26
- FrontierMath (legacy)
- FrontierMath v2 (Tier 4)
- FrontierMath v2 (Tiers 1-3)
- HMMT Feb 2025
- HMMT Feb 2026
- HMMT Nov 2025
- IMOAnswerBench
- MATH-500
- MMAnswerBench
Knowledge & reasoning
- AA-GPQA Diamond
- AA-HLE
- ARC-AGI-2
- ARC-AGI-1
- ARC-AGI-3
- Artificial Analysis Intelligence Index
- BioMysteryBench (human-difficult)
- BioMysteryBench (human-solvable)
- C-Eval
- CritPt
- GPQA
- GPQA Diamond
- HealthBench (raw)
- HealthBench Hard
- HealthBench (length-adjusted)
- HealthBench Professional
- HealthBench Professional (raw)
- HLE
- HLE w/o tools
- HLE-Verified
- JevBench 1.4
- LABBench2
- MedXpertQA (Text)
- MMLU
- MMLU-Pro
- MMLU-Pro (Arcee)
- MMLU-Redux
- MMMLU
- AA-Omniscience Accuracy
- SuperGPQA
- Vals GPQA Diamond
- Vals MMLU-Pro
Instruction following
Multilingual
Vision & documents
- AA-MMMU-Pro
- BabyVision
- CharXiv
- CharXiv w/o tools
- ERQA
- MathVision
- MathVision w/ Python
- MedXpertQA (MM)
- MMMU
- MMMU-Pro
- MMMU-Pro w/ Python
- MMVU
- OfficeQA Pro
- OmniDocBench 1.5
- RealWorldQA
- RefCOCO (avg)
- ScreenSpot Pro
- SimpleVQA
- Video-MME (with subtitle)
- VideoMMMU
- V*
- ZeroBench