HyperAI
Command Palette
Search for a command to run...
GPQA Diamond
GPQA 中最具挑战性的 198 道题，博士级专家可达到 65% 的准确率，而熟练的非专家即使可以使用网络也仅达到 34%。
- 模型
- 205 当前，617 总共
- 中位数
- 71.3%
- 范围
- 11.1%–96.3%
| 排名 | 比较 | 模型 | 评分 |
|---|---|---|---|
| 151 |  Olmo 3.1 32B Instruct Allen Institute for AI | 53.9% | |
| 152 |  Hermes 4 - Llama-3.1 405B (Non-reasoning) Nous Research | 53.6% | |
| 153 |  Reka Flash 3 Reka AI | 52.9% | |
| 154 |  Command A Cohere | 52.7% | |
| 155 | Olmo 3 7B Think Allen Institute for AI | 51.6% | |
| 156 |  Exaone 4.0 1.2B (Reasoning) LG AI Research | 51.5% | |
| 157 | 51.5% | ||
| 158 |  LFM2.5-8B-A1B Liquid AI | 51.3% | |
| 159 | 51.3% | ||
| 160 | 49.8% | ||
| 161 | Hermes 4 - Llama-3.1 70B (Non-reasoning) Nous Research | 49.1% | |
| 162 | LFM2 24B A2B Liquid AI | 47.4% | |
| 163 |  Ministral 3 8B Mistral | 47.1% | |
| 164 | 46.9% | ||
| 165 |  Qwen3.5 2B (Reasoning) Alibaba | 45.6% | |
| 166 | 43.9% | ||
| 167 |  G9v3-3B AI9Stars | 43.8% | |
| 168 | Qwen3.5 2B (Non-reasoning) Alibaba | 43.8% | |
| 169 |  Gemma 4 E2B (Reasoning) Google | 43.3% | |
| 170 | 43.2% | ||
| 171 | Molmo2-8B Allen Institute for AI | 42.5% | |
| 172 | Exaone 4.0 1.2B (Non-reasoning) LG AI Research | 42.4% | |
| 173 | 41.6% | ||
| 174 | 41.2% | ||
| 175 | 40.5% | ||
| 176 | Olmo 3 7B Instruct Allen Institute for AI | 40.0% | |
| 177 | 39.9% | ||
| 178 |  Jamba 1.7 Large AI21 Labs | 39.0% | |
| 179 | DeepHermes 3 - Mistral 24B Preview (Non-reasoning) Nous Research | 38.2% | |
| 180 | Ministral 3 3B Mistral | 35.8% | |
| 181 |  Nova Micro Amazon | 35.8% | |
| 182 | LFM2.5-1.2B-Thinking Liquid AI | 33.9% | |
| 183 | 33.6% | ||
| 184 | Jamba Reasoning 3B AI21 Labs | 33.3% | |
| 185 |  Phi-4 Mini Instruct Microsoft | 33.1% | |
| 186 | LFM2.5-1.2B-Instruct Liquid AI | 32.6% | |
| 187 | Jamba 1.7 Mini AI21 Labs | 32.2% | |
| 188 | Phi-4 Multimodal Instruct Microsoft | 31.5% | |
| 189 |  MiniCPM-V 4.6 1.3B OpenBMB | 30.5% | |
| 190 | Tiny Aya Global Cohere | 30.5% | |
| 191 |  K2 Horizon 0.9B Institute of Foundation Models | 29.3% | |
| 192 | LFM2.5-VL-1.6B Liquid AI | 28.9% | |
| 193 | MiniCPM5-1B (Reasoning) OpenBMB | 27.8% | |
| 194 |  Apertus 70B Instruct Swiss AI Initiative | 27.2% | |
| 195 | DeepHermes 3 - Llama-3.1 8B Preview (Non-reasoning) Nous Research | 27.0% | |
| 196 | MiniCPM5-1B (Non-reasoning) OpenBMB | 26.9% | |
| 197 | 26.3% | ||
| 198 | 26.1% | ||
| 199 | 25.7% | ||
| 200 | Apertus 8B Instruct Swiss AI Initiative | 25.6% |
