AGI-Eval Community - Cognition: leaderboard
Metric: Accuracy (%). Source: agi-eval.cn. 141 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Claude Fable 5 | 88.43 |
| 2 | Gemini 3.1 Pro (Preview) | 88.09 |
| 3 | Claude Opus 4.6 | 87.64 |
| 4 | Gemini 3.5 Flash | 87.43 |
| 5 | Gemini 3 Pro (Preview) | 87.33 |
| 6 | Gemini 3 Flash (Preview) | 86.91 |
| 7 | Claude Opus 4.8 | 86.37 |
| 8 | Claude Opus 4.5 (Non-reasoning) | 86.15 |
| 9 | GPT-5.5 | 86.06 |
| 10 | Claude Opus 4.5 | 86.05 |
| 11 | Grok 4 | 86.02 |
| 12 | Kimi K3 | 85.74 |
| 13 | Qwen 3.7 Max | 85.44 |
| 14 | Seed 2.1 Pro | 85.07 |
| 15 | Kimi K2.6 | 84.87 |
Interactive version: theaggregate.ai/benchmark?slug=agi-eval-community-cognition · How It Works · Data refreshed daily, snapshot 2026-09-19.