AGI-Eval Community - Learning: leaderboard
Metric: Accuracy (%). Source: agi-eval.cn. 141 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Gemini 3.1 Pro (Preview) | 94.78 |
| 2 | Gemini 3 Flash (Preview) | 93.89 |
| 3 | Gemini 3.5 Flash | 92.41 |
| 4 | Gemini 3 Pro (Preview) | 92.27 |
| 5 | GLM-5.1 | 92.26 |
| 6 | Qwen 3.7 Max | 92.03 |
| 7 | GPT-5.5 | 91.89 |
| 8 | GLM-5.2 | 91.86 |
| 9 | DeepSeek V4 Pro (Max) | 91.55 |
| 10 | GLM-5 | 91.4 |
| 11 | Claude Opus 4.6 | 91.22 |
| 12 | Claude Opus 4.8 | 91.13 |
| 13 | GLM-4.7 | 91.02 |
| 14 | Qwen 3.5 397B A17B (Thinking) | 90.99 |
| 15 | Kimi K3 | 90.9 |
Interactive version: theaggregate.ai/benchmark?slug=agi-eval-community-learning · How It Works · Data refreshed daily, snapshot 2026-09-19.