ATBench - Failure Mode Diagnosis: leaderboard
ATBench (2026) tests LLMs and guard models as safety judges of long-horizon agent tool-use trajectories built from a taxonomy of risk sources, failure modes and real-world harms. This board is the accuracy in percent of naming the failure mode of each unsafe trajectory; higher is better.
Metric: Fine-grained diagnosis accuracy (%) of the failure-mode label (ATBench taxonomy) over the 497 unsafe ATBench trajectories, with the AgentDoG diagnosis template; higher is better. Source: arxiv.org. Saturation forecast: Around 2034. 13 models tracked. Updated 7 Oct 2026.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | AgentDoG-Qwen3-4B | 16.5 | |
| 2 | GPT-5.4 | 13.5 | #138 |
| 3 | Gemini 3.1 Pro (Preview) | 12.6 | #102 |
| 4 | GPT-5.2 | 12 | #190 |
| 5 | Qwen 3 235B A22B 2507 Instruct | 11.6 | #625 |
| 6 | Qwen 3 4B 2507 Instruct | 9.6 | #2240 |
| 7 | QwQ-32B | 9.4 | #913 |
| 8 | Gemini 3 Flash | 8.3 | #180 |
| 9 | Qwen 3 4B | 8.2 | #2379 |
| 10 | Qwen 2.5 7B Instruct | 6 | #2732 |
| 11 | Llama 3.1 8B Instruct | 5.8 | #4655 |
| 12 | Qwen 3.5 397B A17B | 3.6 | #270 |
| 13 | Qwen 3.5 4B | 3 | #1270 |
No result here: #2 Claude Fable 5.2, #3 Gemini 4 Neon, #4 GPT-6.2 Astra.
Interactive version: theaggregate.ai/benchmark?slug=atbench-failure-mode-diagnosis · How It Works · Data refreshed daily, snapshot 2026-11-22.