OpenCompass Agent - Tool Use (CompassBench 2507): leaderboard
Metric: Score (%). Source: rank.opencompass.org.cn. 19 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Doubao-Seed-1.6-thinking-250715 | 68.7 |
| 2 | GPT-4.1 (2025-04-14) | 67.4 |
| 3 | Qwen 3 235B A22B 2507 Instruct | 66.8 |
| 4 | Kimi K2 | 66.5 |
| 5 | MiniMax M1 80k | 64.7 |
| 6 | DeepSeek V3 (0324) | 62.6 |
| 7 | GPT-5 | 57.2 |
| 8 | Claude Sonnet 4 (Thinking) | 55.6 |
| 9 | ERNIE-X1-Turbo-32K | 45.9 |
| 10 | Hunyuan-T1-20250711 | 39.1 |
| 11 | GLM 4.5 Air | 34.3 |
| 12 | GLM-4.5 | 34.3 |
| 13 | MiniMax-Text-01 | 30.7 |
Interactive version: theaggregate.ai/benchmark?slug=opencompass-agent-tool-use-compassbench-2507 · How It Works · Data refreshed daily, snapshot 2026-09-19.