ToolMaze (Failure-Aware Prompt) - 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, a failure-aware prompt that warns of tool failures and outlines recovery strategies (the clean runs keep the standard prompt); higher is better. Source: arxiv.org. Saturation forecast: Around March 2027. 9 models tracked.

Top models

#ModelScore
1Gemini 3.1 Pro (Preview)46.25
2GPT-5.543.56
3Claude Sonnet 4.642
4DeepSeek V4 Pro40.81
5GLM-5.133.75
6Qwen 3.6 27B23.94
7Qwen 3.5 397B A17B22.06
8MiniMax-M2.721.94
9Qwen 3.5 35B A3B19.31

Interactive version: theaggregate.ai/benchmark?slug=toolmaze-failure-aware-prompt-perturbed-task-success · How It Works · Data refreshed daily, snapshot 2026-09-29.