AGI-Eval Community - Subject Reasoning: leaderboard
Metric: Accuracy (%). Source: agi-eval.cn. 141 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Claude Opus 4.8 | 90.01 |
| 2 | GPT-5.5 | 89.61 |
| 3 | Claude Fable 5 | 89.25 |
| 4 | Seed 2.0 Pro | 89.25 |
| 5 | Gemini 3.5 Flash | 88.92 |
| 6 | Seed 2.1 Pro | 88.88 |
| 7 | Qwen 3.7 Max | 88.84 |
| 8 | Kimi K3 | 88.84 |
| 9 | Claude Opus 4.6 | 88.81 |
| 10 | Gemini 3.1 Pro (Preview) | 88.63 |
| 11 | Gemini 3 Pro (Preview) | 88.55 |
| 12 | Kimi K2.6 | 88.55 |
| 13 | Kimi K2.5 (Thinking) | 88.01 |
| 14 | DeepSeek V4 Flash (Max) | 87.97 |
| 15 | Claude Opus 4.5 | 87.93 |
Interactive version: theaggregate.ai/benchmark?slug=agi-eval-community-subject-reasoning · How It Works · Data refreshed daily, snapshot 2026-09-19.