OpenCompass Math - Competition - Chinese (CompassBench 2501): leaderboard
OpenCompass CompassBench 2501 edition (file dated 2025-02-20): competition mathematics sub-task of the math category, Chinese questions. CompassBench is OpenCompass's periodic evaluation of chat models on a private question set; each edition is a separate board. Score 0-100, higher is better.
Metric: Score (%). Source: rank.opencompass.org.cn. Saturation forecast: Estimated already saturated. 40 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | DeepSeek R1 | 55 | #262 |
| 2 | O3 Mini (2025-01-31) | 55 | #249 |
| 3 | O1 (2024-12-17) | 50 | #157 |
| 4 | O1 Mini (2024-09-12) | 40 | #412 |
| 5 | GLM-Zero-Preview | 30 | #682 |
| 6 | Gemini 2.0 Pro (Preview 02-05) | 25 | #318 |
| 7 | QwQ 32B-Preview | 25 | #671 |
| 8 | DeepSeek V3 | 20 | #329 |
| 9 | Mistral Large 2 (Nov) Instruct (2411) | 20 | #480 |
| 10 | Llama 3.1 405B Instruct FP8 | 20 | #400 |
| 11 | Llama 3.1 8B Instruct | 15 | #1017 |
| 12 | Claude 3.5 Sonnet (20241022) | 15 | #302 |
| 13 | Qwen 2.5 7B Instruct | 15 | #863 |
| 14 | GPT-4o (2024-11-20) | 15 | #392 |
| 15 | Gemini 2.0 Flash (001) | 15 | #365 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=opencompass-math-competition-chinese-compassbench-2501 · How It Works · Data refreshed daily, snapshot 2026-11-05.