Benchmarks / Agentic
Toolathlon-Verified (HKUST NLP)
An agent completes 108 real-world tool-use tasks across 32 MCP servers and 7 local toolkits, and each task is graded by executing the resulting state against the task's ground truth. Maintained by HKUST NLP. Score: Pass@1 in percent, the mean over three runs (Verified series).
Toolathlon-Verified (HKUST NLP)
Higher is better · best effort per model
full table below
| # | Model | Effort | Score |
|---|---|---|---|
| 1 | GLM 5.3 Flash | max | 78.4% |
| 2 | Kimi K3 | max | 76.5% |
| 3 | Opus 4.8 | max | 76.2% |
| 4 | Muse Spark 1.2 | xhigh | 75.9% |
| 5 | Muse Spark 1.1 | xhigh | 75.6% |
| 6 | DeepSeek V4 Pro 0813 | max | 74.4% |
| 7 | GPT-5.5 | xhigh | 73.5% |
| 8 | Sonnet 5 | max | 71.6% |
| 9 | DeepSeek V4 Flash 0731 | max | 70.7% |
| 10 | Gemini 3.5 Flash | high | 67.3% |
| 11 | Gemini 3.1 Pro | high | 61.1% |
| 12 | GLM 5.2 | max | 59.9% |
| 13 | Kimi K2.6 | max | 58.0% |
| 14 | Kimi K2.7 Code | max | 58.0% |
| 15 | Gemini 3.5 Flash-Lite | high | 57.1% |
| 16 | Hy3 | high | 56.8% |
| 17 | DeepSeek V4 Pro | max | 55.9% |
| 18 | Inkling Small | xhigh | 54.4% |
| 19 | DeepSeek V4 Flash | max | 50.9% |
| 20 | MiMo V2.5 | xhigh | 49.1% |
| 21 | MiniMax M2.7 | high | 47.5% |
| 22 | Inkling | xhigh | 45.5% |
| 23 | Qwen 3.5 397B-A17B | max | 40.7% |
| 24 | Nemotron 3 Ultra 550B A55B | max | 34.3% |
| 25 | Kimi K2.5 | max | 33.0% |