Document Arena
Professional Work · 2026-06-10
Arena-style leaderboard for document analysis and long-content reasoning tasks, reported as arena/Elo-style model scores.
Top models (higher is better)
| Model | Score |
|---|---|
| Claude Opus 5 | 1520 |
| Opus 4.6 | 1510 |
| Claude Fable 5 | 1504 |
| Opus 4.7 | 1498 |
| GPT-5.5 | 1485 |
| Sonnet 4.6 | 1483 |
| GPT-5.6 Terra | 1479 |
| GPT-5.6 Sol | 1479 |
| Opus 4.8 | 1475 |
| Muse Spark 1.1 | 1472 |
| Sonnet 5 | 1470 |
| GPT-5.4 | 1470 |
| Gemini 3.5 Flash | 1465 |
| GPT-5.6 Luna | 1462 |
| Opus 4.5 | 1462 |
| Grok 4.5 | 1454 |
| Kimi K2.6 | 1451 |
| Sonnet 4.5 | 1446 |
| Gemini 3.1 Pro Preview | 1445 |
| Muse Spark | 1443 |
| Qwen3.7-Plus | 1440 |
| MiniMax M3 | 1435 |
| Gemini 3 Pro Preview | 1433 |
| Kimi K2.5 | 1429 |
| Gemma 4 31B | 1424 |
| Haiku 4.5 | 1423 |
| Gemini 2.5 Pro | 1421 |
| GLM 5V Turbo | 1417 |
| Gemini 3 Flash Preview | 1413 |
| GPT-5.2 | 1405 |
| GPT-5.1 | 1401 |