OpenCompass Math - Arithmetic (CompassBench 2405): leaderboard
OpenCompass CompassBench 2405 edition (file dated 2024-06-12): numerical calculation sub-task of the math category. 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. 39 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Claude 3 Opus | 82.7 | #565 |
| 2 | DeepSeek V2 Chat | 80.7 | #572 |
| 3 | GPT-4o (2024-05-13) | 76.3 | #337 |
| 4 | GPT-4 Turbo | 75 | #417 |
| 5 | Mixtral 8x22B Instruct | 74 | #738 |
| 6 | Yi-1.5-9B Chat | 73.7 | #1032 |
| 7 | Command-R+ | 73 | #869 |
| 8 | Yi 1.5 34B Chat | 71 | #798 |
| 9 | GPT-3.5 Turbo | 70.7 | #858 |
| 10 | Llama 3 70B Instruct | 70 | #640 |
| 11 | Yi Large | 69.7 | #566 |
| 12 | Moonshot v1 8K | 69 | #622 |
| 13 | Qwen 1.5 110B Chat | 68.7 | #693 |
| 14 | Mistral Large | 68 | #713 |
| 15 | DeepSeek-V2-Lite-Chat | 66.7 | #1189 |
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-arithmetic-compassbench-2405 · How It Works · Data refreshed daily, snapshot 2026-11-05.