MazeBench ASCII Rooms Visited
Games · 2026-07-28
Exploration score from the same Python-disabled MazeBench ASCII runs as the headline: the share of the world's 255 non-starting rooms the agent visited, as the leaderboard displays it. It measures exploration within the same run rather than puzzle mastery, so it is kept as a diagnostic.
Top models (higher is better)
| Model | Score |
|---|---|
| GPT-6 Astra | 40.4 |
| GPT-6.1 Sol | 25.5 |
| Claude Fable 5.1 | 4.3 |
| Claude Sonnet 5.5 | 2.7 |
| GPT-6 Sol | 2.4 |
| Grok 4.7 | 2.4 |
| Claude Haiku 5.5 | 1.6 |
| Muse Wonder | 1.2 |
| Muse Spark 1.3 | 0.8 |
| GPT-6 Luna | 0.4 |
| Muse Spark 1.2 | 0.0 |