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 |
|---|---|
| Claude Fable 5 | 53.3 |
| Claude Opus 5 | 52.6 |
| GPT-5.6 Sol | 47.2 |
| Opus 4.8 | 45.7 |
| Muse Spark 1.1 | 45.1 |
| Gemini 3.1 Pro Preview | 44.7 |
| Kimi K3 | 44.3 |
| GPT-5.5 | 44.3 |
| GPT-5.6 Terra | 41.8 |
| GPT-5.4 | 41.6 |
| Gemini 3.5 Flash | 41.0 |
| Grok 4.5 | 40.3 |
| GLM-5.2 | 40.1 |
| Muse Spark | 39.9 |
| GPT-5.3-Codex | 39.9 |
| Opus 4.7 | 39.6 |
| Sonnet 5 | 39.6 |
| Gemini 3.6 Flash | 38.3 |
| Motif-3-Beta | 38.2 |
| Qwen3.7-Max | 38.1 |
| Gemini 3 Pro Preview | 37.2 |
| GPT-5.6 Luna | 37.2 |
| MiniMax M3 | 37.1 |
| DeepSeek-V4-Flash-0731 | 36.8 |
| Opus 4.6 | 36.7 |
| Grok Build 0.1 | 36.0 |
| Kimi K2.6 | 35.9 |
| DeepSeek-V4-Pro | 35.9 |
| GPT-5.2 | 35.4 |
| Grok 4.3 | 35.0 |
| Gemini 3 Flash Preview | 34.7 |
| MiMo-V2.5-Pro | 33.8 |
| GPT-5.2-Codex | 33.5 |
| KAT-Coder-Pro V1 | 33.4 |
| Qwen3.7-Plus | 33.4 |
| Kimi K2.7 Code | 32.8 |
| Nex-N2-Pro | 32.4 |
| Grok 4.20 | 32.2 |
| DeepSeek-V4-Flash | 32.1 |
| LongCat-2.0 | 32.1 |