ToolRobustBench - Runtime Environment: leaderboard
Metric: Success rate (0-1, native; perturbed runtime status and feedback during execution; over ToolRobustBench single-family perturbation instances (clean, light, medium, heavy); binary per-instance tool-call success). Source: arxiv.org. Saturation forecast: Around December 2026. 7 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | DeepSeek V4 Pro | 0.75 |
| 2 | Claude Sonnet 4.6 | 0.75 |
| 3 | Qwen Plus | 0.71 |
| 4 | GPT-5.4 Mini | 0.71 |
| 5 | Gemini 2.5 Flash | 0.65 |
| 6 | Gemini 2.5 Pro | 0.64 |
| 7 | DeepSeek V4 Flash | 0.62 |
Interactive version: theaggregate.ai/benchmark?slug=toolrobustbench-runtime-environment · How It Works · Data refreshed daily, snapshot 2026-09-26.