OpenCompass Agent - Hard Multi-Turn (CompassBench 2510): leaderboard
OpenCompass CompassBench 2510 edition (file dated 2025-11-10): hard multi-turn tool calling sub-task of the agent 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: Around February 2027. 27 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | O4 Mini (2025-04-16) (High) | 58 | #183 (O4 Mini (2025-04-16)) |
| 2 | Qwen 3 Next 80B A3B (Thinking) | 56 | #312 |
| 3 | Claude Sonnet 4.5 (Thinking) | 54 | #148 (Claude Sonnet 4.5) |
| 4 | MiniMax-M2 | 54 | #327 |
| 5 | Gemini 2.5 Pro | 52 | #159 |
| 6 | GLM-4.6 | 52 | #265 |
| 7 | Kimi K2 0905 | 52 | #292 |
| 8 | Claude Haiku 4.5 (Thinking) | 52 | #287 (Claude Haiku 4.5) |
| 9 | DeepSeek R1 0528 | 50 | #232 |
| 10 | O3 (2025-04-16) (High) | 50 | #125 (O3 (2025-04-16)) |
| 11 | DeepSeek V3.2 Exp | 48 | #241 |
| 12 | Doubao-Seed-1.6-251015 (Thinking) | 48 | #208 (Doubao-Seed-1.6-251015) |
| 13 | Qwen 3 235B A22B 2507 (Thinking) | 44 | #263 |
| 14 | GPT-OSS-120B (High) | 43.8 | #346 (GPT-OSS-120B) |
| 15 | Seed-OSS-36B-Instruct | 40 | #384 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=opencompass-agent-hard-multi-turn-compassbench-2510 · How It Works · Data refreshed daily, snapshot 2026-11-05.