OpenCompass Math - Arithmetic (CompassBench 2409): leaderboard
OpenCompass CompassBench 2409 edition (file dated 2024-10-09): 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. 30 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Qwen 2.5 72B Instruct | 88.3 | #469 |
| 2 | Claude 3.5 Sonnet (20240620) | 86 | #308 |
| 3 | GPT-4o (2024-08-06) | 85.3 | #343 |
| 4 | GPT-4o (2024-05-13) | 84.7 | #337 |
| 5 | DeepSeek V2.5 | 82.7 | #494 |
| 6 | Qwen 2.5 7B Instruct | 79.7 | #863 |
| 7 | Mistral Large 2 (Jul) | 79.3 | #490 |
| 8 | ERNIE-4.0-Turbo-8K-Preview | 78.7 | #603 |
| 9 | Gemini 1.5 Pro | 78.3 | #454 |
| 10 | Llama 3.1 405B Instruct FP8 | 76.7 | #400 |
| 11 | GPT-4o Mini (2024-07-18) | 76.3 | #568 |
| 12 | GLM-4 Plus | 76 | #398 |
| 13 | Llama 3.1 70B Instruct | 74 | #567 |
| 14 | Mistral-Small-Instruct-2409 | 72.3 | #709 |
| 15 | Step 2 16K | 72.3 | #442 |
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-2409 · How It Works · Data refreshed daily, snapshot 2026-11-05.