Reasoning
Humanitys-Last-Exam
El benchmark mas dificil conocido - problemas academicos novedosos.
27 modelos publicaron score
| # | Modelo | Empresa | Score |
|---|---|---|---|
| 1 | Claude Mythos 5 | Anthropic | 59.0 |
| 2 | Muse Spark | Meta | 58.0 |
| 3 | Claude Mythos Preview | Anthropic | 56.8 |
| 4 | Claude Opus 4.7 | Anthropic | 54.7 |
| 5 | Claude Fable 5 | Anthropic | 53.0 |
| 6 | Grok 4 Heavy | xAI | 50.7 |
| 7 | Kimi K2.5 | Moonshot AI | 50.2 |
| 8 | Claude Opus 4.8 | Anthropic | 49.8 |
| 9 | Gemini 3 Deep Think | Google DeepMind | 48.4 |
| 10 | Gemini 3.1 Pro | Google DeepMind | 44.4 |
| 11 | GLM-4.7 | Zhipu AI | 42.8 |
| 12 | Qwen3.7-Max | Alibaba | 41.4 |
| 13 | GLM-5.2 | Zhipu AI | 40.5 |
| 14 | Gemini 3.5 Flash | Google DeepMind | 40.2 |
| 15 | Claude Opus 4.6 | Anthropic | 40.0 |
| 16 | DeepSeek V4 Pro | DeepSeek | 37.7 |
| 17 | Gemini 3 Pro | Google DeepMind | 37.5 |
| 18 | Kimi K2.6 | Moonshot AI | 34.7 |
| 19 | Gemini 3 Flash | Google DeepMind | 33.7 |
| 20 | GLM-5.1 | Zhipu AI | 31.0 |
| 21 | GLM-5 | Zhipu AI | 30.5 |
| 22 | Nemotron 3 Ultra 550B-A55B | Nvidia | 26.7 |
| 23 | Grok 4 | xAI | 25.4 |
| 24 | MiMo V2 Flash | Xiaomi | 22.1 |
| 25 | Nemotron 3 Super | Nvidia | 18.3 |
| 26 | Gemini 3.1 Flash-Lite | Google DeepMind | 16.0 |
| 27 | K-EXAONE 236B-A23B | LG AI Research | 13.6 |