HyperAI
Command Palette
Search for a command to run...
Humanity's Last Exam
一项前沿级基准，包含 2,500 道经专家审核的数学、科学和人文学科题目，旨在作为最终的封闭式学术评测。
- 模型
- 267 当前，684 总共
- 中位数
- 12.0%
- 范围
- 1.1%–74.3%
| 排名 | 比较 | 模型 | 评分 |
|---|---|---|---|
| 251 | 3.8% | ||
| 252 | 3.8% | ||
| 253 |  Llama 4 Scout Meta | 3.8% | |
| 254 |  DeepHermes 3 - Mistral 24B Preview (Non-reasoning) Nous Research | 3.8% | |
| 255 |  Phi-4 Microsoft | 3.8% | |
| 256 |  Gemma 4 E4B (Reasoning) Google | 3.8% | |
| 257 |  Jamba Reasoning 3B AI21 Labs | 3.8% | |
| 258 | Jamba 1.7 Large AI21 Labs | 3.7% | |
| 259 | Gemma 3 270M Google | 3.6% | |
| 260 |  K2-V2 (Low) Institute of Foundation Models | 3.6% | |
| 261 | Hermes 4 - Llama-3.1 70B (Non-reasoning) Nous Research | 3.6% | |
| 262 | 3.6% | ||
| 263 |  ERNIE 4.5 300B A47B Baidu | 3.3% | |
| 264 | 2.9% | ||
| 265 |  Qwen3.5 2B (Reasoning) Alibaba | 2.6% | |
| 266 | 2.5% | ||
| 267 | Qwen3.5 0.8B (Reasoning) Alibaba | 1.1% |
