ATM-Bench Hard — leaderboard
Metric: Oracle QS (%). Source: atmbench.github.io. 18 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | GPT-5.5 | 71.49 |
| 2 | GPT-5 | 66.05 |
| 3 | GPT-5.4 | 65.19 |
| 4 | Gemini 2.5 Pro | 64.3 |
| 5 | Claude Opus 4.5 | 62.7 |
| 6 | Qwen 3.6 27B | 62.3 |
| 7 | Claude Sonnet 4.5 | 61.9 |
| 8 | MiniMax-M3 | 61.8 |
| 9 | Gemini 3 Flash | 60.1 |
| 10 | Gemini 2.5 Flash | 55.6 |
| 11 | Gemini 3.1 Pro (Preview) | 55.4 |
| 12 | Gemini 3 Pro | 52.1 |
| 13 | MiMo-V2.5 | 52.1 |
| 14 | GPT-5.2 | 50.41 |
| 15 | Gemini 3.1 Flash Lite | 45.7 |
Interactive version: theaggregate.ai/benchmark?slug=atm-bench-hard · How It Works · Data refreshed daily, snapshot 2026-08-05.