Arena AI Document — leaderboard
Crowdsourced Arena AI pairwise human-preference leaderboard for PDF and document-understanding models.
Metric: Arena ELO (self-reported). Source: benchmarklist.com. Status: saturation imminent. 19 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Claude Opus 4.6 | 1526 |
| 2 | Claude Opus 4.7 | 1515 |
| 3 | Claude Sonnet 4.6 | 1500 |
| 4 | GPT-5.5 | 1490 |
| 5 | GPT-5.4 | 1480 |
| 6 | Claude Opus 4.5 | 1470 |
| 7 | Muse Spark | 1457 |
| 8 | Claude Sonnet 4.5 | 1450 |
| 9 | Gemini 3.1 Pro (Preview) | 1449 |
| 10 | Gemma 4 31B | 1432 |
| 11 | Gemini 2.5 Pro | 1429 |
| 12 | Grok 4.20 | 1426 |
| 13 | Claude Haiku 4.5 | 1424 |
| 14 | Gemini 3 Flash (Preview) | 1421 |
| 15 | GPT-5.2 (High) | 1414 |
Interactive version: theaggregate.ai/benchmark?slug=arena-ai-document · How the rankings work · Data refreshed daily, snapshot 2026-07-22.