AgentNoiseBench (tau2-bench, Tool Noise): leaderboard
Metric: Success on a stratified 25 percent sample of tau2-bench tasks (airline and retail) with tool noise injected into tool returns (execution failures, incomplete, erroneous, misleading or redundant outputs); noise produced by a fixed generator whose prompt was optimized to degrade a reference agent while keeping every task solvable; task success rate (fraction times 100), mean of four trials; the paper's stability-gated criterion counts a task solved only with a correct final outcome and no noise-induced deviation in the trajectory; higher is better. Source: arxiv.org. Saturation forecast: Around May 2027. 24 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-5.2 (Thinking) | 61 | #105 (GPT-5.2) |
| 2 | Claude Sonnet 4.5 (Thinking) | 57 | #138 (Claude Sonnet 4.5) |
| 3 | O3 | 57 | #121 |
| 4 | Qwen 3 Max (Thinking) | 56 | #201 (Qwen 3 Max) |
| 5 | GLM-4.6 (Thinking) | 56 | #246 (GLM-4.6) |
| 6 | GLM-4.5 (Thinking) | 49 | #265 (GLM-4.5) |
| 7 | Claude Sonnet 4.5 | 48 | #138 |
| 8 | GLM-4.6 (Non-reasoning) | 48 | #246 (GLM-4.6) |
| 9 | LongCat-Flash-Chat | 48 | #322 |
| 10 | Claude Sonnet 4 (Thinking) | 46 | #194 (Claude Sonnet 4) |
| 11 | GLM-4.5 (Non-reasoning) | 42 | #265 (GLM-4.5) |
| 12 | Claude Sonnet 4 | 41 | #194 |
| 13 | Doubao-Seed-1.6 (Thinking) | 40 | |
| 14 | Gemini 2.5 Pro | 37 | #145 |
| 15 | DeepSeek R1 0528 | 37 | #217 |
No result here: #3 Claude Opus 5.5, #5 GPT-6 Astra, #8 Claude Fable 5.1.
Interactive version: theaggregate.ai/benchmark?slug=agentnoisebench-tau2-bench-tool-noise · How It Works · Data refreshed daily, snapshot 2026-10-11.