OpenCompass Reasoning - Puzzle - Chinese (CompassBench 2501): leaderboard
Metric: Score (%). Source: rank.opencompass.org.cn. 40 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | O3 Mini (2025-01-31) | 92 |
| 2 | DeepSeek R1 | 91.2 |
| 3 | O1 (2024-12-17) | 91.2 |
| 4 | Doubao-1.5-pro-32k-250115 | 86.4 |
| 5 | O1 Mini (2024-09-12) | 85.6 |
| 6 | Gemini 2.0 Flash (001) | 82.4 |
| 7 | Gemini 2.0 Pro (Preview 02-05) | 78.4 |
| 8 | Claude 3.5 Sonnet (20241022) | 77.6 |
| 9 | DeepSeek V3 | 76 |
| 10 | GLM-Zero-Preview | 76 |
| 11 | QwQ 32B-Preview | 75.2 |
| 12 | Qwen 2.5 Max | 75.2 |
| 13 | GPT-4o (2024-11-20) | 69.6 |
| 14 | Qwen 2.5 72B Instruct | 68.8 |
| 15 | GLM-4 Plus | 68.8 |
Interactive version: theaggregate.ai/benchmark?slug=opencompass-reasoning-puzzle-chinese-compassbench-2501 · How It Works · Data refreshed daily, snapshot 2026-09-19.