Ebr bench — Epoch AI
Ebr bench — Epoch AI is run by Epoch AI. It has ranked 21 models, of which the catalogue holds 21, scoring from 2.381 to 76.19.
measured by Epoch AI · the board itself
| Place | Model | Metric | Score |
|---|---|---|---|
| 1stof 21 | GPT 6 Astra | mean_score (max) | 76.19 |
| 2ndof 21 | Claude Fable 5.1 | mean_score (max) | 57.143 |
| 3rdof 21 | Claude Opus 5 | mean_score (max) | 45.714 |
| 4thof 21 | GPT-5.6 Sol | mean_score (max) | 44.762 |
| 5thof 21 | Claude Fable 5 | mean_score (max) | 39.524 |
| 6thof 21 | GPT-5.5 | mean_score (xhigh) | 34.286 |
| 7thof 21 | Grok 4.6 | mean_score (xhigh) | 30.476 |
| 8thof 21 | Claude Opus 4.8 | mean_score (max) | 28.571 |
| 9thof 21 | GPT-5.4 | mean_score (xhigh) | 25.397 |
| 10thof 21 | GPT-5.2 | mean_score (xhigh) | 23.016 |
| 11thof 21 | Claude Opus 4.7 | mean_score (max) | 19.048 |
| 12thof 21 | Claude Opus 4.5 | mean_score | 14.286 |
| 13thof 21 | Gemini 3.1 Pro Preview | mean_score | 14.286 |
| 14thof 21 | Claude Opus 4.6 | mean_score (max) | 12.698 |
| 15thof 21 | GPT-5 | mean_score (high) | 12.698 |
| 16thof 21 | GLM 5.2 | mean_score (max) | 9.524 |
| 17thof 21 | Qwen3.7 Max | mean_score (max) | 9.524 |
| 18thof 21 | Claude Opus 4.1 | mean_score | 7.937 |
| 19thof 21 | Gemini 3.5 Flash | mean_score (high) | 4.762 |
| 20thof 21 | Kimi K2.6 | mean_score | 2.381 |
| 21stof 21 | Claude Sonnet 4.5 | mean_score | 2.381 |
Read from the board on 2026-09-16