OpenCompass LLM - Instruction Following (CompassBench 2501): leaderboard
OpenCompass CompassBench 2501 edition (file dated 2025-02-20): instruction-following category score. 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 | 74 | #262 |
| 2 | GPT-4o (2024-11-20) | 67 | #392 |
| 3 | Gemini 2.0 Pro (Preview 02-05) | 64.8 | #318 |
| 4 | Yi Lightning | 64 | #429 |
| 5 | MiniMax-Text-01 | 63.2 | #519 |
| 6 | GLM-4 Plus | 62.9 | #398 |
| 7 | Gemini 2.0 Flash (001) | 62.5 | #365 |
| 8 | DeepSeek V3 | 58 | #329 |
| 9 | GPT-4o Mini (2024-07-18) | 53.5 | #568 |
| 10 | Qwen 2.5 Max | 52.7 | #427 |
| 11 | Mistral Large 2 (Nov) Instruct (2411) | 52.4 | #480 |
| 12 | Qwen 2.5 72B Instruct | 48.8 | #469 |
| 13 | Yi 1.5 34B Chat | 48.3 | #798 |
| 14 | Claude 3.5 Sonnet (20241022) | 48.2 | #302 |
| 15 | QwQ 32B-Preview | 43.5 | #671 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=opencompass-llm-instruction-following-compassbench-2501 · How It Works · Data refreshed daily, snapshot 2026-11-05.