Humanity's Last Exam (Epoch AI replication)
Humanity's Last Exam (Epoch AI replication) is run by Epoch AI (AI Benchmarking Hub). It has ranked 56 models, of which the catalogue holds 41, scoring from 2.72 to 54.8 on Accuracy.
measured by Epoch AI (AI Benchmarking Hub) · the board itself
| Place | Model | Metric | Score |
|---|---|---|---|
| 1stof 56 | GPT 6 Astra | Accuracy | 54.8 |
| 2ndof 56 | Claude Fable 5.1 | Accuracy (xhigh) | 46.5 |
| 3rdof 56 | Gemini 3.1 Pro Preview | Accuracy | 46.44 |
| 4thof 56 | Claude Sonnet 5.5 | Accuracy (xhigh) | 46.12 |
| 5thof 56 | Gemini 3.8 Flash | Accuracy | 44.52 |
| 6thof 56 | GPT-5.4 Pro | Accuracy | 44.32 |
| 7thof 56 | Muse Spark 1.2 | Accuracy | 40.56 |
| 8thof 56 | gemini-3-pro-preview | Accuracy | 37.52 |
| 9thof 56 | GPT-5.4 | Accuracy (xhigh) | 36.24 |
| 10thof 56 | Claude Opus 4.7 | Accuracy | 36.2 |
| 11thof 56 | Claude Opus 4.6 | Accuracy (max) | 34.44 |
| 12thof 56 | GPT-5 Pro | Accuracy | 31.64 |
| 13thof 56 | Claude Haiku 5.5 | Accuracy | 30.12 |
| 14thof 56 | GPT-5.2 | Accuracy | 27.8 |
| 15thof 56 | GPT-5 | Accuracy (high) | 25.32 |
| 17thof 56 | Claude Opus 4.5 | Accuracy | 25.2 |
| 18thof 56 | Kimi K2.5 | Accuracy | 24.37 |
| 19thof 56 | GPT-5.1 | Accuracy | 23.68 |
| 20thof 56 | Gemini 2.5 Pro Preview 06-05 | Accuracy | 21.64 |
| 21stof 56 | o3 | Accuracy (high) | 20.32 |
| 22ndof 56 | GPT-5 mini | Accuracy | 19.44 |
| 26thof 56 | o4-mini | Accuracy (high) | 18.08 |
| 27thof 56 | Gemini 2.5 Pro Preview 05-06 | Accuracy | 17.8 |
| 30thof 56 | Claude Sonnet 4.5 | Accuracy | 13.72 |
| 31stof 56 | Gemini 2.5 Flash | Accuracy | 12.08 |
| 32ndof 56 | Claude Opus 4.1 | Accuracy | 11.52 |
| 34thof 56 | Claude Opus 4 | Accuracy | 10.72 |
| 35thof 56 | Gemini 3.1 Flash-Lite | Accuracy | 8.64 |
| 36thof 56 | GLM-4.5 | Accuracy | 8.32 |
| 37thof 56 | GLM-4.5-Air | Accuracy | 8.12 |
| 38thof 56 | o1-pro | Accuracy | 8.12 |
| 39thof 56 | Claude 3.7 Sonnet | Accuracy | 8.04 |
| 40thof 56 | o1 | Accuracy | 7.96 |
| 42ndof 56 | Claude Sonnet 4 | Accuracy | 7.76 |
| 47thof 56 | Llama-4-Maverick-17B-128E-Instruct | Accuracy | 5.68 |
| 50thof 56 | GPT-4.1 fine-tuned | Accuracy | 5.4 |
| 52ndof 56 | Mistral Medium 2505 | Accuracy | 4.52 |
| 53rdof 56 | Nova Pro | Accuracy | 4.4 |
| 54thof 56 | Claude 3.5 Sonnet | Accuracy | 4.08 |
| 55thof 56 | Nova Lite | Accuracy | 3.64 |
| 56thof 56 | GPT-4o (2024-11-20) | Accuracy | 2.72 |
Read from the board on 2026-11-01