OpenCompass LLM - Math (CompassBench 2504): leaderboard
Metric: Score (%). Source: rank.opencompass.org.cn. 28 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | O4 Mini (High) | 68.8 |
| 2 | O3 Mini (High) | 67.6 |
| 3 | Gemini 2.5 Pro (Preview 05-06) | 66 |
| 4 | Qwen 3 32B (Thinking) | 62.9 |
| 5 | DeepSeek R1 | 61.8 |
| 6 | Doubao-1.5-thinking-pro-250415 | 59 |
| 7 | Qwen 3 235B A22B (Thinking) | 58.3 |
| 8 | ERNIE-X1-Turbo-32K | 55.9 |
| 9 | Grok 3 | 53 |
| 10 | DeepSeek V3 (0324) | 50 |
| 11 | GPT-4.1 (2025-04-14) | 45.1 |
| 12 | GPT-4.1 Mini | 43.2 |
| 13 | GLM-4 32B (0414) | 42.2 |
| 14 | Qwen 3 235B A22B | 38.7 |
| 15 | Llama 4 Maverick Instruct | 35.4 |
Interactive version: theaggregate.ai/benchmark?slug=opencompass-llm-math-compassbench-2504 · How It Works · Data refreshed daily, snapshot 2026-09-19.