Coding
SWE-bench Verified leaderboard
As of 2026-11-21, the highest measured score on SWE-bench Verified is 96.0% by Claude Opus 5. 172 more models have estimated scores, calibrated from the benchmarks they were measured on.
Measured scores: benchlm.ai.
| # | Model | Score | Source |
|---|---|---|---|
| 1 | Claude Fable 5.2 | 100.0% | estimated ± 3.6 pp, medium confidence |
| 2 | Claude Fable 5.1 | 97.8% | estimated ± 3.6 pp, medium confidence |
| 3 | Claude Opus 5 | 96.0% | measured |
| 4 | Claude Mythos 5 | 95.5% | measured |
| 5 | Claude Fable 5 | 95.0% | measured |
| 6 | Claude Opus 5.5 | 94.5% | estimated ± 2.2 pp, low confidence |
| 7 | GPT-6 Astra | 93.9% | estimated ± 2.2 pp, medium confidence |
| 8 | Muse Wonder Full | 92.3% | estimated ± 5.1 pp, low confidence |
| 9 | Muse Wonder | 92.3% | estimated ± 5.1 pp, low confidence |
| 10 | SWE-2 | 92.2% | estimated ± 2.2 pp, medium confidence |
| 11 | Ember-1 | 92.2% | measured |
| 12 | Sakana Fugu-Ultra | 91.8% | estimated ± 3.6 pp, high confidence |
| 13 | Qwen4 Max | 90.8% | estimated ± 5.1 pp, low confidence |
| 14 | Kimi K3.1 Pro | 90.8% | estimated ± 5.1 pp, low confidence |
| 15 | Muse Spark 1.3 | 90.7% | estimated ± 5.1 pp, low confidence |
| 16 | Hy4 | 90.5% | estimated ± 5.1 pp, medium confidence |
| 17 | Claude Haiku 5.5 | 90.4% | estimated ± 2.2 pp, medium confidence |
| 18 | Claude Sonnet 5.5 | 90.3% | estimated ± 2.2 pp, medium confidence |
| 19 | Gemini 4 Argon | 90.2% | estimated ± 4.5 pp, medium confidence |
| 20 | Pareto 26.9 | 90.0% | estimated ± 5.1 pp, medium confidence |
| 21 | SWE-2 Flash | 89.9% | estimated ± 2.2 pp, medium confidence |
| 22 | Grok 4.6 | 89.9% | estimated ± 4.4 pp, high confidence |
| 23 | MAI Code 2 | 89.7% | estimated ± 5.1 pp, medium confidence |
| 24 | Particle-1 | 89.4% | estimated ± 5.1 pp, medium confidence |
| 25 | GPT-6.1 Sol | 89.0% | estimated ± 5.1 pp, medium confidence |
| 26 | MiMo-V2.6-Pro | 89.0% | estimated ± 5.1 pp, medium confidence |
| 27 | Gemini 3.7 Flash | 89.0% | estimated ± 2.2 pp, medium confidence |
| 28 | Kimi K3 | 88.7% | estimated ± 4.4 pp, high confidence |
| 29 | GPT-5.5 | 88.7% | estimated ± 2.2 pp, medium confidence |
| 30 | Grok 4.7 | 88.6% | estimated ± 5.1 pp, medium confidence |
| 31 | Claude Opus 4.8 | 88.6% | measured |
| 32 | SWE-1.7 | 88.3% | estimated ± 2.2 pp, medium confidence |
| 33 | Pareto 26.10 Preview | 88.1% | estimated ± 5.1 pp, medium confidence |
| 34 | Ling 3.1 Flash | 87.8% | estimated ± 3.6 pp, high confidence |
| 35 | GPT-6.1 Terra | 87.8% | estimated ± 6.3 pp, medium confidence |
| 36 | Claude Opus 4.7 | 87.6% | measured |
| 37 | GPT-6 Sol | 87.5% | estimated ± 5.1 pp, medium confidence |
| 38 | Mistral Large 4 | 87.3% | estimated ± 4.5 pp, medium confidence |
| 39 | MiMo-V2.6-Flash | 87.1% | estimated ± 5.1 pp, medium confidence |
| 40 | Step 5 Preview | 87.0% | estimated ± 5.1 pp, medium confidence |
| 41 | GPT-6 Luna | 86.5% | estimated ± 5.1 pp, medium confidence |
| 42 | Ornith-1.5-397B | 86.0% | measured |
| 43 | Hy4 flash | 85.6% | estimated ± 5.1 pp, medium confidence |
| 44 | Muse Spark 1.2 | 85.3% | estimated ± 4.4 pp, high confidence |
| 45 | Claude Sonnet 5 | 85.2% | measured |
| 46 | GPT-5.3 Codex | 85.0% | measured |
| 47 | Grok 4.5 | 84.5% | estimated ± 3.6 pp, high confidence |
| 48 | GPT-5.6 Sol | 84.5% | estimated ± 3.6 pp, high confidence |
| 49 | Ultramarine 1 Flash | 84.2% | estimated ± 5.1 pp, medium confidence |
| 50 | Muse Spark Open | 83.6% | estimated ± 5.1 pp, medium confidence |
| 51 | MAI Code 2 Flash | 83.6% | estimated ± 5.1 pp, medium confidence |
| 52 | GPT-5.6 Terra | 83.5% | estimated ± 3.6 pp, high confidence |
| 53 | Qwen3.8 Max Preview | 83.4% | estimated ± 6.3 pp, medium confidence |
| 54 | Composer 2 Fast | 83.2% | estimated ± 5.0 pp, low confidence |
| 55 | Gemini 3.1 Pro | 82.9% | estimated ± 3.2 pp, medium confidence |
| 56 | GPT-5.6 Luna | 82.9% | estimated ± 3.6 pp, high confidence |
| 57 | Ornith-1.0-397B | 82.4% | measured |
| 58 | GPT-5.4 | 82.4% | estimated ± 3.2 pp, medium confidence |
| 59 | DeepSeek V4.1 Flash 1017 | 82.3% | estimated ± 2.3 pp, medium confidence |
| 60 | Inkling 2 Small | 82.2% | measured |
| 61 | ERNIE 5.5 Flash Preview | 82.2% | estimated ± 2.3 pp, medium confidence |
| 62 | GPT-5.1-Codex-Max | 82.1% | estimated ± 3.2 pp, medium confidence |
| 63 | Muse Spark 1.1 | 82.0% | estimated ± 3.6 pp, high confidence |
| 64 | MiniMax M3.1 Flash | 82.0% | estimated ± 2.3 pp, high confidence |
| 65 | LongCat-2.5 | 81.9% | estimated ± 2.3 pp, high confidence |
| 66 | Gemini 3.8 Flash | 81.9% | estimated ± 4.4 pp, high confidence |
| 67 | Hy4 preview | 81.8% | estimated ± 2.3 pp, high confidence |
| 68 | GLM-5.3 | 81.7% | estimated ± 2.3 pp, high confidence |
| 69 | Sage-A1.5 | 81.7% | estimated ± 3.6 pp, high confidence |
| 70 | Composer 2.5 | 81.7% | estimated ± 4.4 pp, high confidence |
| 71 | Gemini 3.6 Flash | 81.7% | estimated ± 4.4 pp, high confidence |
| 72 | Beam 1.0 501B-A23B | 81.6% | measured |
| 73 | GLM-5.3-Flash | 81.6% | estimated ± 2.3 pp, high confidence |
| 74 | Qwen3.8 Max | 81.6% | estimated ± 2.3 pp, high confidence |
| 75 | Gemini 3 Pro | 81.5% | estimated ± 3.2 pp, medium confidence |
| 76 | o3 | 81.4% | estimated ± 13.1 pp, low confidence |
| 77 | GPT-5.2-Codex | 81.1% | estimated ± 3.2 pp, medium confidence |
| 78 | Claude Opus 4.6 (Adaptive) | 80.9% | estimated ± 4.4 pp, high confidence |
| 79 | Kimi K2.7 Code | 80.9% | estimated ± 4.4 pp, high confidence |
| 80 | Claude Opus 4.5 | 80.9% | measured |
| 81 | Claude Opus 4.6 | 80.8% | measured |
| 82 | GLM-5.2 | 80.7% | estimated ± 2.3 pp, high confidence |
| 83 | Qwen3.8-Omni-Flash | 80.7% | estimated ± 2.3 pp, high confidence |
| 84 | Beam 1.0 501B-A23B NVFP4 | 80.7% | measured |
| 85 | Qwen3.8-Flash-Next | 80.6% | estimated ± 2.3 pp, high confidence |
| 86 | DeepSeek V4 Pro | 80.6% | measured |
| 87 | MiniMax M3 | 80.5% | measured |
| 88 | GPT-5.4 mini | 80.5% | estimated ± 2.2 pp, medium confidence |
| 89 | Atria Dawn Preview | 80.4% | estimated ± 3.6 pp, high confidence |
| 90 | Seed 2.1 Pro | 80.4% | estimated ± 2.3 pp, high confidence |
| 91 | Qwen3.7 Max | 80.4% | measured |
| 92 | Laguna S 2.1 | 80.3% | estimated ± 3.6 pp, high confidence |
| 93 | Inkling-Small | 80.2% | measured |
| 94 | Kimi K2.6 | 80.2% | measured |
| 95 | Beam 1.0 501B-A23B FP8 | 80.2% | measured |
| 96 | Claude Opus 4.5 Thinking | 80.0% | estimated ± 4.4 pp, high confidence |
| 97 | GPT-5.2 | 80.0% | measured |
| 98 | Sakana Fugu | 80.0% | estimated ± 3.6 pp, high confidence |
| 99 | Mercury 3 | 80.0% | estimated ± 5.1 pp, medium confidence |
| 100 | Seed 2.1 Turbo | 79.7% | estimated ± 2.3 pp, high confidence |
| 101 | Claude Sonnet 4.6 | 79.6% | measured |
| 102 | Qwen 3.6 Max (preview) | 79.5% | estimated ± 2.3 pp, high confidence |
| 103 | GLM-5.1 | 79.5% | estimated ± 2.3 pp, high confidence |
| 104 | Qwen3.8-27B | 79.4% | estimated ± 2.3 pp, high confidence |
| 105 | Gemini 3 Flash | 79.3% | estimated ± 4.4 pp, high confidence |
| 106 | Inkling 2 Mini | 79.2% | measured |
| 107 | DeepSeek V4 Flash 0731 | 79.0% | measured |
| 108 | Ornith-1.5-35B-A3B | 79.0% | measured |
| 109 | Qwen3.6 Plus | 78.8% | measured |
| 110 | MiniMax M2.7 | 78.6% | estimated ± 2.3 pp, high confidence |
| 111 | o4-mini | 78.6% | estimated ± 13.1 pp, low confidence |
| 112 | MiMo-V2.5-Pro | 78.5% | estimated ± 3.6 pp, high confidence |
| 113 | BTL-4 | 78.4% | measured |
| 114 | dots3-note Preview | 78.4% | measured |
| 115 | Hy3 | 78.4% | estimated ± 6.3 pp, medium confidence |
| 116 | Ling 3.0 Flash | 78.0% | estimated ± 3.6 pp, high confidence |
| 117 | DeepSeek V3.2 | 78.0% | estimated ± 4.8 pp, medium confidence |
| 118 | MiMo-V2-Pro | 78.0% | measured |
| 119 | GLM-5 | 77.8% | measured |
| 120 | Step 3.7 Flash | 77.8% | estimated ± 3.6 pp, high confidence |
| 121 | Apodex 1.1 | 77.7% | measured |
| 122 | Qwen3.7 Plus | 77.7% | measured |
| 123 | Quasar 438B | 77.6% | estimated ± 6.3 pp, medium confidence |
| 124 | MiMo-V2.5 | 77.6% | estimated ± 3.6 pp, high confidence |
| 125 | Inkling | 77.6% | measured |
| 126 | Mistral Medium 3.5 128B | 77.6% | measured |
| 127 | Grok 4.3 | 77.4% | estimated ± 4.4 pp, high confidence |
| 128 | Muse Spark | 77.4% | measured |
| 129 | Claude Sonnet 4.5 | 77.2% | measured |
| 130 | Qwen3.6-27B | 77.2% | measured |
| 131 | Gemini 3.5 Flash | 76.8% | estimated ± 3.6 pp, high confidence |
| 132 | Kimi K2.5 | 76.8% | measured |
| 133 | Claude Sonnet 4.5 Thinking | 76.7% | estimated ± 4.4 pp, high confidence |
| 134 | Grok 4.20 | 76.7% | measured |
| 135 | GPT-5.1 | 76.6% | estimated ± 4.4 pp, high confidence |
| 136 | GPT-5.4 nano | 76.6% | estimated ± 4.4 pp, high confidence |
| 137 | Ripple-65B Preview | 76.5% | measured |
| 138 | GPT-5 | 76.2% | estimated ± 4.4 pp, high confidence |
| 139 | Qwen3.5 397B | 76.2% | measured |
| 140 | Gemma 4 31B | 76.1% | estimated ± 5.0 pp, medium confidence |
| 141 | Gemini 3.5 Flash-Lite | 76.1% | estimated ± 3.6 pp, high confidence |
| 142 | Muse Glimmer 30B | 76.0% | measured |
| 143 | Composer 2 | 75.7% | estimated ± 4.8 pp, medium confidence |
| 144 | Ornith-1.0-35B | 75.6% | measured |
| 145 | MiMo-V2-Omni | 74.8% | measured |
| 146 | Laguna M.1 | 74.6% | measured |
| 147 | Claude 4.1 Opus | 74.5% | measured |
| 148 | Hy3 Preview | 74.4% | measured |
| 149 | Step 3.5 Flash | 74.4% | measured |
| 150 | Flint | 74.2% | measured |
| 151 | Qwen3.5 Plus | 74.2% | estimated ± 4.5 pp, high confidence |
| 152 | MiniMax M2.5 | 74.0% | estimated ± 4.5 pp, high confidence |
| 153 | Apodex 1.1 Mini | 73.9% | estimated ± 6.3 pp, medium confidence |
| 154 | Qwen3.5 Flash | 73.9% | estimated ± 4.4 pp, high confidence |
| 155 | GPT-5 mini | 73.9% | estimated ± 4.5 pp, high confidence |
| 156 | GLM-4.7 | 73.8% | measured |
| 157 | GPT-5.1-Codex | 73.6% | estimated ± 4.5 pp, high confidence |
| 158 | BTL-3 | 73.5% | estimated ± 9.7 pp, low confidence |
| 159 | MAI-Thinking-1 | 73.5% | measured |
| 160 | MiMo-V2-Flash | 73.4% | measured |
| 161 | Qwen3.6-35B-A3B | 73.4% | measured |
| 162 | Claude Haiku 4.5 | 73.3% | measured |
| 163 | Claude Haiku 4.5 Thinking | 73.3% | estimated ± 4.5 pp, high confidence |
| 164 | Gemini 3.1 Flash-Lite | 73.0% | estimated ± 4.4 pp, high confidence |
| 165 | Jais-3-MoVA-36B-A4B | 72.8% | measured |
| 166 | Claude 4 Sonnet | 72.7% | measured |
| 167 | MAI-Code-1.1-Flash | 72.6% | measured |
| 168 | Claude 4 Opus | 72.5% | measured |
| 169 | Jais-3-32B | 72.4% | measured |
| 170 | Qwen3.5-27B | 72.4% | measured |
| 171 | Ling 3.0 Flash VL | 72.0% | estimated ± 6.3 pp, medium confidence |
| 172 | Qwen3.5-122B-A10B | 72.0% | measured |
| 173 | Nemotron 3 Ultra | 71.9% | measured |
| 174 | Qwen3 Max | 71.6% | estimated ± 4.5 pp, high confidence |
| 175 | GLM-4.6 | 71.5% | estimated ± 4.5 pp, high confidence |
| 176 | Nemotron 3.5 Super VL 120B A12B | 71.1% | measured |
| 177 | Nemotron 3.5 Super VL 120B A12B NVFP4 | 71.1% | measured |
| 178 | Laguna XS 2.1 | 70.9% | measured |
| 179 | Grok Code Fast 1 | 70.8% | measured |
| 180 | Ornith-1.5-9B | 70.6% | measured |
| 181 | Solar Pro 4 | 70.6% | measured |
| 182 | Grok 4 | 70.5% | estimated ± 4.4 pp, high confidence |
| 183 | Solar Open 2 | 70.4% | measured |
| 184 | Jais-3-375B-A23B | 70.2% | measured |
| 185 | Jais-3-7B | 70.2% | measured |
| 186 | Laguna XS.2 | 69.9% | measured |
| 187 | Qwen3-Coder 480B-A35B | 69.6% | measured |
| 188 | A.X K2 | 69.5% | estimated ± 5.6 pp, medium confidence |
| 189 | Ornith-1.0-9B | 69.4% | measured |
| 190 | Qwen3.5-35B-A3B | 69.2% | measured |
| 191 | Ling 3.0 Flash FP8 | 68.4% | estimated ± 5.6 pp, medium confidence |
| 192 | K-EXAONE 2.0 | 68.2% | measured |
| 193 | LongCat-Flash-Lite-Sparse | 68.2% | measured |
| 194 | Trinity-Large-Preview | 66.8% | estimated ± 6.3 pp, medium confidence |
| 195 | Trinity-Large-Thinking | 66.8% | estimated ± 6.3 pp, medium confidence |
| 196 | o1 | 66.2% | estimated ± 6.5 pp, medium confidence |
| 197 | DeepSeek V3.1 | 66.0% | measured |
| 198 | Gemma 4 26B A4B | 65.9% | estimated ± 6.3 pp, medium confidence |
| 199 | Kimi K2 | 65.8% | measured |
| 200 | Nemotron 3.5 Bolt 74B-A9B | 65.5% | estimated ± 6.3 pp, medium confidence |
| 201 | DeepSeek V3 0324 | 64.5% | estimated ± 6.3 pp, medium confidence |
| 202 | Mistral Small 4 | 64.2% | estimated ± 6.3 pp, medium confidence |
| 203 | North Mini Code | 64.2% | estimated ± 6.3 pp, medium confidence |
| 204 | Grok 4 Fast (Reasoning) | 64.1% | estimated ± 4.4 pp, medium confidence |
| 205 | Mercury 2.5 | 64.0% | estimated ± 5.6 pp, medium confidence |
| 206 | Grok 3 Mini | 63.8% | estimated ± 9.3 pp, low confidence |
| 207 | Gemini 2.5 Pro | 63.8% | measured |
| 208 | Command A+ | 63.8% | estimated ± 6.3 pp, medium confidence |
| 209 | Spark-X2.6-4B | 62.8% | measured |
| 210 | GPT-OSS 120B | 62.4% | measured |
| 211 | Claude 3.7 Sonnet | 62.3% | measured |
| 212 | Grok 4.1 Fast (Reasoning) | 62.0% | estimated ± 4.4 pp, medium confidence |
| 213 | Mistral Large 3 | 62.0% | estimated ± 4.4 pp, medium confidence |
| 214 | Ternary Bonsai 2 27B | 60.8% | measured |
| 215 | o1-preview | 60.8% | estimated ± 6.5 pp, medium confidence |
| 216 | GPT-OSS 20B | 60.7% | measured |
| 217 | Nemotron Ultra 253B | 60.5% | estimated ± 9.3 pp, low confidence |
| 218 | Nemotron 3 Super 100B | 60.5% | measured |
| 219 | GLM-4.7-Flash | 59.2% | measured |
| 220 | DeepSeek-R1-0528-Qwen3-8B | 58.0% | estimated ± 9.3 pp, low confidence |
| 221 | DeepSeek-R1-0528 | 57.6% | measured |
| 222 | Mellum2-12B-A2.5B-Thinking | 57.5% | estimated ± 9.7 pp, low confidence |
| 223 | ZAYA1-8B | 57.5% | estimated ± 9.7 pp, low confidence |
| 224 | Qwen3 235B 2507 | 57.5% | estimated ± 9.7 pp, low confidence |
| 225 | Magistral Medium | 57.5% | estimated ± 9.7 pp, low confidence |
| 226 | Magistral Small | 57.5% | estimated ± 9.7 pp, low confidence |
| 227 | LLaDA2.2-mini | 57.5% | estimated ± 9.7 pp, low confidence |
| 228 | Mellum2-12B-A2.5B-Instruct | 57.5% | estimated ± 9.7 pp, low confidence |
| 229 | MiniCPM5-1B | 57.5% | estimated ± 9.7 pp, low confidence |
| 230 | Mistral Medium 3 | 57.5% | estimated ± 9.7 pp, low confidence |
| 231 | Gemma 4 12B | 57.5% | estimated ± 6.5 pp, medium confidence |
| 232 | GLM-4.5 | 57.2% | estimated ± 13.1 pp, low confidence |
| 233 | Granite 4.2 30B | 57.0% | measured |
| 234 | DeepSeek R1 Distill Llama 70B | 56.6% | estimated ± 9.3 pp, low confidence |
| 235 | Grok 3 [Beta] | 56.3% | estimated ± 9.3 pp, low confidence |
| 236 | MiniMax M1 80k | 56.0% | measured |
| 237 | Phi-4-reasoning | 54.7% | estimated ± 9.3 pp, low confidence |
| 238 | GPT-4.1 | 54.6% | measured |
| 239 | DeepSeek R1 Distill Qwen 14B | 54.4% | estimated ± 9.3 pp, low confidence |
| 240 | Phi-4-reasoning-plus | 54.4% | estimated ± 9.3 pp, low confidence |
| 241 | Llama 4 Maverick | 54.0% | estimated ± 6.3 pp, medium confidence |
| 242 | ZAYA1-74B-Preview | 53.2% | measured |
| 243 | Nemotron 3.5 Lightning 30B A3B NVFP4 | 52.8% | measured |
| 244 | Nemotron 3 Nano Omni 30B A3B | 51.8% | estimated ± 5.6 pp, medium confidence |
| 245 | LFM2.5-24B-A2B | 51.0% | measured |
| 246 | K-Exaone | 49.4% | measured |
| 247 | o3-mini | 49.3% | measured |
| 248 | LLaDA2.2-flash | 49.3% | measured |
| 249 | DeepSeek-R1 | 49.2% | measured |
| 250 | Claude 3.5 Sonnet | 49.0% | measured |
| 251 | Claude 3.7 Sonnet Thinking | 48.7% | estimated ± 13.1 pp, low confidence |
| 252 | Gemini 1.5 Pro | 48.5% | estimated ± 6.5 pp, medium confidence |
| 253 | Mistral Small 3.1 | 48.3% | estimated ± 6.3 pp, medium confidence |
| 254 | Granite 4.2 8B | 47.7% | measured |
| 255 | Agents-A1-4B | 47.5% | estimated ± 5.6 pp, medium confidence |
| 256 | Mellum2.1-12B-A2.5B-Thinking | 47.0% | measured |
| 257 | Devstral Small | 46.8% | measured |
| 258 | MiniCPM5-2B | 46.4% | measured |
| 259 | Qwen2.5-Max | 45.7% | estimated ± 9.3 pp, low confidence |
| 260 | GPT-4 Turbo | 45.5% | estimated ± 6.5 pp, medium confidence |
| 261 | Solar Pro 3 | 45.0% | estimated ± 6.3 pp, low confidence |
| 262 | Sarvam 105B | 45.0% | measured |
| 263 | Ling 2.6 Flash | 43.7% | estimated ± 5.6 pp, medium confidence |
| 264 | Gemma 4 E4B | 43.4% | estimated ± 6.3 pp, low confidence |
| 265 | LFM3-1.7B | 43.4% | measured |
| 266 | Ling 3.0 Tiny | 43.1% | estimated ± 6.3 pp, low confidence |
| 267 | Claude 3 Opus | 42.6% | estimated ± 6.5 pp, low confidence |
| 268 | DeepSeek V3 | 42.0% | measured |
| 269 | Gemma 3 27B | 41.8% | estimated ± 6.3 pp, low confidence |
| 270 | Granite 4.2 3B | 40.8% | estimated ± 5.6 pp, low confidence |
| 271 | Claude 3.5 Haiku | 40.6% | measured |
| 272 | Celeris-1 | 39.4% | estimated ± 6.3 pp, low confidence |
| 273 | Llama 4 Scout | 38.9% | estimated ± 6.3 pp, low confidence |
| 274 | Nemotron 3 Nano 30B | 34.0% | estimated ± 6.5 pp, low confidence |
| 275 | Sarvam 30B | 34.0% | measured |
| 276 | Gemini 2.5 Flash-Lite | 31.6% | measured |
| 277 | GPT-4o | 31.2% | estimated ± 13.1 pp, low confidence |
| 278 | Ultravox v0.6 Llama 3.3 70B | 29.4% | estimated ± 6.5 pp, low confidence |
| 279 | LFM2.5-2.6B | 29.0% | estimated ± 6.3 pp, low confidence |
| 280 | GPT-4o mini | 28.3% | estimated ± 6.5 pp, low confidence |
| 281 | GPT-4.1 nano | 27.9% | estimated ± 6.5 pp, low confidence |
| 282 | GPT-4.1 mini | 23.6% | measured |
| 283 | Gemini 2.0 Flash | 21.4% | measured |
| 284 | Gemma 4 E2B | 19.4% | estimated ± 6.5 pp, low confidence |
| 285 | Gemini 2.0 Flash-Lite | 12.5% | measured |
| 286 | Gemini 1.5 Flash | 9.6% | measured |
All leaderboards
Agentic · terminal
- AA Terminal-Bench 2.1
- AA Terminal-Bench 4.0
- Terminal-Bench 4.0
- Terminal-Bench 2.0
- Terminal-Bench 2.1
- Terminal-Bench 3.0
- Terminal-Bench Hard
- Terminal-Bench-Science 0.1
- Terminal-Bench 4.1
- Terminal-Bench 2.1 (Vals AI)
Agentic · tools
- AA Agentic Index
- AA-AnalystAgent
- AA AutomationBench
- CWE-Bench-AA
- AA Cyber Index
- DeepsecBench-AA
- AA EnterpriseOps-Gym
- GDP.pdf
- AA Harvey LAB v1.0
- Harvey LAB-AA v1.1
- AA ITBench
- AA Tau3 Banking
- Agents' Last Exam
- APEX-Agents
- APEX-Agents-AA
- ApprenticeBench
- AutomationBench
- BFCL v4
- BrowseComp
- Claw-Eval
- CWE-bench v1
- CWE-bench
- CyberGym
- DeepPlanning
- DeepSearchQA
- DRACO
- ExploitGym
- Finance Agent v2
- GDPval-AA
- Gert Labs
- Hermes Bench
- Hermes Index
- HLE w/ tools
- JobBench
- LifeVal
- MazeBench
- MCP Atlas
- MCP-Tasks
- MKA
- OSWorld 2.1
- OSWorld 2.1 Offline
- OSWorld-Verified
- OSWorld 2.0
- PinchBench
- QwenClawBench
- ResearchClawBench
- SkillsBench
- τ²-bench results
- τ³-bench results
- Toolathlon
- Toolathlon-Verified
- VITA-Bench
- WideResearch
Coding
- AA Coding Index
- AA LiveCodeBench
- AA-SciCode
- Bug Hunt Bench
- CursorBench 3.1
- CursorBench 3.2
- CursorBench 4.0
- cursorBench41
- DeepSWE
- FrontierCode 1.1 Main
- FrontierCode 2.0
- FrontierCode 1.1 Extended
- FrontierSWE v2
- LiveCodeBench
- LiveCodeBench Pro
- LiveCodeBench v6
- NL2Repo
- OpenHarmony Bench v1.2
- OpenHarmony Bench v1.0
- PostTrainBench
- PostTrainBench v1
- PostTrainBench v1.1
- ProgramBench
- React Native Evals
- ScarfBench
- SciCode
- SWE-Atlas Codebase QnA
- SWE-bench Pro
- SWE-bench Verified
- SWE-Rebench
- SWE-Marathon
- Vals LiveCodeBench
- Vals SWE-bench
- Vibe Code Bench
- VulcanBench Frontier v4
- VulcanBench v3
Math
- AIME 2024
- AIME 2025
- AIME26
- 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
- HealthBench Professional (length-adjusted, Artificial Analysis)
- HealthBench Professional (raw, Artificial Analysis)
- AA-HLE
- AA MMLU-Pro
- ARC-AGI-2
- ARC-AGI-3 Provider Adapter
- ARC-AGI-1
- ARC-AGI-3
- Artificial Analysis Intelligence Index
- BBH
- BioBasicsBench
- 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
- SimpleQA
- STEMBench
- SuperGPQA
- Vals GPQA Diamond
- Vals MMLU-Pro
Instruction following
Multilingual
Vision & documents
- AA-MMMU-Pro
- AI2D_TEST
- BabyVision
- BabyVision w/ Python
- BenchCAD Vision2Code (no tools)
- BenchCAD Vision2Code (tools)
- CC-OCR
- Chartography (no tools)
- Chartography (tools)
- CharXiv
- CharXiv w/o tools
- CountBench
- DynaMath
- ERQA
- LVBench
- MathVision
- MathVision w/ Python
- MedXpertQA (MM)
- MLVU (M-Avg)
- MMMU
- MMMU-Pro
- MMMU-Pro w/ Python
- MMVU
- OCRBench V2
- OfficeQA
- OfficeQA Pro
- OmniDocBench 1.5
- ParseBench-1
- ParseBench-2
- PerceptionBench
- RealWorldQA
- RefCOCO (avg)
- ScreenSpot Pro
- SimpleVQA
- SWE-bench Multimodal
- Video-MME (w/o subtitle)
- Video-MME (with subtitle)
- VideoMMMU
- V*
- We-Math
- ZeroBench
- ZeroBench w/ Python
Long context
General
- HLE-MultiChoice (Decision Index)
- Artificial Analysis Decision Index v1.0
- Artificial Analysis Decision Index
- When2Call (Decision Index)
- WildGuardTest (Decision Index)
- BabaIsBench
- BinaryAudit
- BlenderWorld (text-only)
- Chess Puzzles
- CompileBench
- Cookie-Bench
- DAYJOB
- DAYJOB Healthcare
- Decision Index Full Score
- EBR-bench
- ExploitBench
- ExploitBench Internal Port
- FACTS Multimodal
- FailureSensorIQ
- FrontierPhysics
- Furniture Assembly
- Decision Index Arts
- Decision Index Knowledge
- Decision Index Language
- Decision Index Retrieval
- Decision Index Tools
- MaxxBench
- MirrorCode
- Mystery Game Puzzles
- NYT Connections
- OTelBench
- SEC-Bench Pro
- SkillsBench official leaderboard (BenchFlow)
- SpireClimb
- StealthMark