Humanity's Last Exam (Text-Only)
General QA · 2025-01-24
The 2,158-question text-only subset of Humanity's Last Exam. Tool availability is a run-level setting recorded on each score, not a separate benchmark identity.
Top models (higher is better)
| Model | Score |
|---|---|
| Gemini 4 Carbon | 60.4 |
| Step 5 Preview | 59.4 |
| Claude Opus 5.5 | 57.5 |
| Gemini 4 Argon | 57.1 |
| Claude Sonnet 5.5 | 55.0 |
| Claude Fable 5.1 | 54.6 |
| Claude Opus 5 | 54.4 |
| Muse Wonder Full | 53.6 |
| GPT-6 Astra | 53.1 |
| GPT-6.1 Sol | 52.6 |
| Kimi K3.1 Pro | 52.5 |
| Claude Fable 5 | 51.5 |
| Muse Wonder | 50.1 |
| MiMo-V2.6-Pro | 49.4 |
| Muse Spark 1.3 | 48.7 |
| MiniMax M3.1 Flash | 48.0 |
| GPT-5.6 Sol | 47.3 |
| Gemini 3.1 Pro Preview | 47.0 |
| GPT-6 Sol | 46.3 |
| GPT-6.1 Terra | 46.2 |
| Muse Spark 1.2 | 45.5 |
| GPT-5.5 | 45.0 |
| Opus 4.8 | 44.8 |
| DeepSeek V4.1 Flash 1017 | 44.0 |
| Muse Spark Open | 43.9 |
| JT-4.1 Flash 236B A21B | 43.6 |
| Ling 3.1 Flash 1019 | 43.5 |
| Grok 4.7 | 43.1 |
| Qwen3.8 Max (0902) | 43.1 |
| Qwen3.8-Max | 43.0 |
| Grok 4.6 | 42.9 |
| Claude Haiku 5.5 | 42.7 |
| Grok 4.5 | 42.7 |
| GPT-5.3-Codex | 42.5 |
| Qwen3.8 2.4T A95B | 42.4 |
| Opus 4.7 | 42.2 |
| Gemini 3.8 Flash | 42.1 |
| GPT-5.6 Terra | 41.9 |
| Gemini 3.5 Flash | 41.3 |
| Muse Spark | 40.7 |