ATBench: 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 F1 score in percent for classifying whole trajectories as safe or unsafe, with unsafe as the positive class; higher is better.
Metric: Binary safe or unsafe classification F1 (%), unsafe as the positive class, on the 1,000 ATBench agent trajectories (503 safe, 497 unsafe; average 9 turns, tools drawn from pools of 2,084), each full trajectory judged with the AgentDoG prompt template (guard models with their native templates); higher is better. Source: arxiv.org. Saturation forecast: Around February 2027. 16 models tracked. Updated 7 Oct 2026.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-5.4 | 76.7 | #135 |
| 2 | Gemini 3.1 Pro (Preview) | 75 | #99 |
| 3 | Gemini 3 Flash | 74.9 | #175 |
| 4 | GPT-5.2 | 71.8 | #185 |
| 5 | AgentDoG-Qwen3-4B | 71.1 | |
| 6 | Qwen 3.5 397B A17B | 67.8 | #264 |
| 7 | Llama 3.1 8B Instruct | 61.9 | #4600 |
| 8 | Qwen 3 235B A22B 2507 Instruct | 60.8 | #616 |
| 9 | Llama Guard 4 12B | 41.7 | |
| 10 | QwQ-32B | 31 | #907 |
| 11 | Qwen 3.5 4B | 27.6 | #1241 |
| 12 | Qwen 3 4B 2507 Instruct | 25.5 | #2221 |
| 13 | Qwen 2.5 7B Instruct | 17.1 | #2723 |
| 14 | Qwen 3 4B | 11.9 | #2348 |
| 15 | Llama Guard 3 8B | 7.3 |
No result here: #1 Claude Fable 5.2, #3 Gemini 4 Neon, #4 GPT-6.2 Astra.
Interactive version: theaggregate.ai/benchmark?slug=atbench · How It Works · Data refreshed daily, snapshot 2026-11-21.