SciArena: leaderboard
SciArena — community arena where researchers vote pairwise on model answers to scientific-literature questions; Elo-style ranking (Allen AI).
Metric: Elo Rating. Source: sciarena.allen.ai. Saturation forecast: Around November 2026. 38 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | O3 | 1151.4 | #124 |
| 2 | Claude Opus 4.1 | 1126.4 | #140 |
| 3 | GPT-5 | 1122.5 | #98 |
| 4 | Gemini 3 Pro (Preview) | 1086.3 | #66 |
| 5 | GPT-5.1 | 1079.7 | #135 |
| 6 | Claude Opus 4 | 1071.7 | #163 |
| 7 | GPT-5 Mini | 1067.1 | #186 |
| 8 | Gemini 2.5 Pro | 1061.3 | #152 |
| 9 | Grok 4 | 1044.6 | #181 |
| 10 | DeepSeek R1 0528 | 1041.8 | #222 |
| 11 | GPT-OSS-120B | 1036.5 | #341 |
| 12 | Qwen 3 235B A22B 2507 (Thinking) | 1036.3 | #260 (Qwen 3 235B A22B 2507) |
| 13 | O4 Mini | 1028.6 | #180 |
| 14 | Claude Sonnet 4 | 1025.2 | #199 |
| 15 | Qwen 3 235B A22B 2507 | 1020.5 | #260 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=sciarena · How It Works · Data refreshed daily, snapshot 2026-10-20.