Benchmarks · reasoning
EQ-Bench 3
Uses the published Elo leaderboard. Score: Elo scaled expected win rate vs claude-fable-5. Collected by Jake Boggs' benchmark dashboard.
70 models · 70 results · data as of 2026-10-29 · eqbench.com ↗
Leaderboard
Fable 5 max
All results
70 rows| # | Model | Effort | EQ-Bench 3 |
|---|---|---|---|
| 1 | Fable 5 | max | 2096 |
| 2 | Opus 4.8 | max | 2058 |
| 3 | Opus 4.7 | max | 1914 |
| 4 | Opus 4.6 | max | 1712 |
| 5 | Sonnet 4.6 | max | 1710 |
| 6 | GLM 5.2 | max | 1563 |
| 7 | DeepSeek V4 Pro | max | 1555 |
| 8 | GPT-5.5 | xhigh | 1554 |
| 9 | Gemini 3 Pro | high | 1553 |
| 10 | Kimi K2.6 | max | 1548 |
| 11 | Kimi K2 | medium | 1548 |
| 12 | GPT-5.2 | xhigh | 1544 |
| 13 | GLM 5.1 | max | 1539 |
| 14 | GPT-5.4 | xhigh | 1537 |
| 15 | Sonnet 4.5 | high | 1535 |
| 16 | Kimi K2.5 | max | 1530 |
| 17 | Gemini 3.1 Pro | high | 1527 |
| 18 | GPT-5.1 | high | 1527 |
| 19 | Gemini 2.5 Pro | high | 1523 |
| 20 | Opus 4.5 | high | 1522 |
| 21 | DeepSeek V4 Flash | max | 1504 |
| 22 | o3 | high | 1500 |
| 23 | GLM 5 | high | 1495 |
| 24 | Gemma 4 31B IT | high | 1464 |
| 25 | GLM 4.7 | none | 1463 |
| 26 | Gemma 4 26B | xhigh | 1455 |
| 27 | Qwen 3.5 397B-A17B | max | 1451 |
| 28 | GPT-5.3 | xhigh | 1446 |
| 29 | Opus 4 | medium | 1440 |
| 30 | Gemma 4 12B IT | 1416 |