# Arena Elo (Documents) — AI model rankings

Real people hand two anonymous AIs the same PDF or document and vote for whichever answers better. The votes become a chess-style Elo rating on arena.ai. Unlike a fixed document benchmark, the files are whatever people actually brought along. Higher is better.

25 tracked models have a published Arena Elo (Documents) score. Higher is better. Scores are as published at each model's release.

## Ranking

| Rank | Model | Developer | Score | Released |
| --- | --- | --- | --- | --- |
| 1 | Claude Opus 5 | Anthropic | 1520 | Jul 24 2026 |
| 2 | Claude Opus 4.6 | Anthropic | 1510 | Feb 5 2026 |
| 3 | Claude Fable 5 | Anthropic | 1504 | Jun 9 2026 |
| 4 | Claude Opus 4.7 | Anthropic | 1498 | Apr 16 2026 |
| 5 | GPT-5.5 | OpenAI | 1485 | Apr 23 2026 |
| 6 | Claude Sonnet 4.6 | Anthropic | 1483 | Feb 17 2026 |
| 7 | GPT-5.6 Terra | OpenAI | 1479 | Jun 26 2026 |
| 7 | GPT-5.6 Sol | OpenAI | 1479 | Jun 26 2026 |
| 9 | Claude Opus 4.8 | Anthropic | 1475 | May 28 2026 |
| 10 | Muse Spark 1.1 | Meta | 1472 | Jul 9 2026 |
| 11 | GPT-5.4 | OpenAI | 1470 | Mar 5 2026 |
| 11 | Claude Sonnet 5 | Anthropic | 1470 | Jun 30 2026 |
| 13 | Gemini 3.5 Flash | Google | 1465 | May 19 2026 |
| 14 | GPT-5.6 Luna | OpenAI | 1462 | Jun 26 2026 |
| 15 | Grok 4.5 | SpaceXAI | 1454 | Jul 8 2026 |
| 16 | Kimi K2.6 | Moonshot AI | 1451 | Apr 21 2026 |
| 17 | Gemini 3.1 Pro | Google | 1445 | Feb 19 2026 |
| 18 | Muse Spark | Meta | 1443 | Apr 8 2026 |
| 19 | Qwen3.7-Plus | Qwen | 1440 | Jun 1 2026 |
| 20 | Gemini 3.0 Pro | Google | 1433 | Nov 18 2025 |
| 21 | Kimi K2.5 | Moonshot AI | 1429 | Jan 27 2026 |
| 22 | Gemini 2.5 Pro | Google | 1421 | Mar 25 2025 |
| 23 | Gemini 3.0 Flash | Google | 1413 | Dec 17 2025 |
| 24 | GPT-5.2 | OpenAI | 1405 | Dec 11 2025 |
| 25 | GPT-5.1 | OpenAI | 1401 | Nov 12 2025 |


---

Canonical page: https://aireleasetracker.com/benchmark/arena-document-elo
Full dataset: https://aireleasetracker.com/llms-full.txt · JSON: https://aireleasetracker.com/models.json
Source: AI Release Tracker (https://aireleasetracker.com). Benchmark scores are the figures published by the releasing lab at launch.