OpenCompass Math - Arithmetic (CompassBench 2407): leaderboard
Metric: Score (%). Source: rank.opencompass.org.cn. 36 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | GPT-4o (2024-05-13) | 87 |
| 2 | Claude 3.5 Sonnet (20240620) | 85.3 |
| 3 | DeepSeek V2 Chat (0628) | 80 |
| 4 | GPT-4o Mini (2024-07-18) | 78.3 |
| 5 | Mistral Large 2 (Jul) | 77 |
| 6 | Llama 3.1 70B Instruct | 74.3 |
| 7 | Mistral Large | 73.7 |
| 8 | Gemma 2 27B (IT) | 70.7 |
| 9 | Mixtral 8x22B Instruct (v0.1) | 69.3 |
| 10 | Yi Large | 66.7 |
| 11 | Gemma 2 9B (IT) | 64.7 |
| 12 | Moonshot v1 8K | 64.3 |
| 13 | GLM-4 (0520) | 64 |
| 14 | DBRX | 56.7 |
| 15 | Llama 3.1 8B Instruct | 56.3 |
Interactive version: theaggregate.ai/benchmark?slug=opencompass-math-arithmetic-compassbench-2407 · How It Works · Data refreshed daily, snapshot 2026-09-19.