Do Agents Know What They Can't Do? Evaluating: leaderboard
Infeasibility-awareness benchmark — can tool-using agents detect that a task is impossible under a constrained tool environment instead of burning compute; self-reported average across settings.
Metric: Avg. (self-reported). Source: benchmarklist.com. Saturation forecast: Around July 2027. 9 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-5.5 | 61.2 | #31 |
| 2 | GPT-OSS-120B | 60.7 | #341 |
| 3 | Qwen 3.5 397B A17B | 58.2 | #146 |
| 4 | Qwen 3.5 27B | 56.7 | #187 |
| 5 | DeepSeek V4 Pro | 54.8 | #102 |
| 6 | DeepSeek V4 Flash | 52.9 | #137 |
| 7 | Qwen 3.5 35B A3B | 52.3 | #256 |
| 8 | Qwen 3.5 122B A10B | 52.3 | #173 |
| 9 | Qwen 3.5 9B | 47.2 | #379 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=do-agents-know-what-they-can-t-do-evaluating · How It Works · Data refreshed daily, snapshot 2026-10-20.