ToolMaze - Perturbed Task Success: leaderboard
Metric: Task success rate (%) averaged over the four tool-failure modes, macro-averaged over the four topology levels; 400 DAG-structured tool-use tasks (four topology levels) over 270 tools, each also run under four injected tool-failure modes (explicit or implicit, transient or permanent); sandboxed, 25 agent steps, temperature 1, the standard tool-use prompt; higher is better. Source: arxiv.org. Saturation forecast: Around September 2028. 9 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | GPT-5.5 | 31.19 |
| 2 | DeepSeek V4 Pro | 31.12 |
| 3 | GLM-5.1 | 27.25 |
| 4 | Gemini 3.1 Pro (Preview) | 26.38 |
| 5 | Claude Sonnet 4.6 | 21.94 |
| 6 | MiniMax-M2.7 | 19 |
| 7 | Qwen 3.5 35B A3B | 16.94 |
| 8 | Qwen 3.6 27B | 16.88 |
| 9 | Qwen 3.5 397B A17B | 16.62 |
Interactive version: theaggregate.ai/benchmark?slug=toolmaze-perturbed-task-success · How It Works · Data refreshed daily, snapshot 2026-09-29.