BabaIsBench Stage 0: The Intro (pass@3)
Games · 2026-07-16
Share of BabaIsBench's eight Stage 0 levels (The Intro of Baba Is You, played through a text simulator of the game's engine in Harbor) that a model solves in at least one of its three Terminus 2 attempts, each limited to 60 minutes. It comes from the same attempts as the pass@1 headline and is a component of its factor group; 24 of the 32 models on the September 5, 2026 board reach 100%.
Top models (higher is better)
| Model | Score |
|---|---|
| Claude Fable 5.1 | 100.0 |
| Claude Fable 5 | 100.0 |
| Opus 4.8 | 100.0 |
| Claude Opus 5 | 100.0 |
| Gemini 3.1 Pro Preview | 100.0 |
| Gemini 3.5 Flash | 100.0 |
| Gemini 3.6 Flash | 100.0 |
| Gemini 3.7 Flash | 100.0 |
| Gemini 3.8 Flash | 100.0 |
| GLM-5.2 | 100.0 |
| GLM-5.3 | 100.0 |
| GPT-5.5 | 100.0 |
| GPT-5.6 Sol | 100.0 |
| GPT-5.6 Terra | 100.0 |
| GPT-6 Astra | 100.0 |
| Grok 4.6 | 100.0 |
| Kimi K3 | 100.0 |
| Sonnet 5 | 88.0 |
| DeepSeek-V4-Flash-0731 | 88.0 |
| DeepSeek V4 Pro 0813 | 88.0 |
| GLM-5.3 Flash | 88.0 |
| GPT-5.6 Luna | 88.0 |
| Grok 4.5 | 88.0 |
| Hy3 | 75.0 |
| Qwen3.8-Max | 75.0 |
| MiniMax M3 | 63.0 |
| Muse Spark 1.3 | 63.0 |
| DeepSeek-V4-Pro | 38.0 |
| Qwen3.6 27B | 38.0 |
| Qwen3.7-Max (source unspecified) | 38.0 |
| Qwen3.8 27B | 38.0 |