SnakeBench: frontier models still fail to plan more than 2-3 moves ahead
GregKamradt · x · 2026-11-20
Greg Kamradt ran Astra 6.2 in SnakeBench against gpt-5.5: models still fail to anticipate more than 2-3 moves ahead, with Astra backing itself into an inescapable corner.
He proposes manufacturing snake scenarios requiring N-step anticipation (1, 2, 3, 4, 5...) to track long-horizon planning improvements. Models play via regular prompting, no codex.
More from Models
- Demis Hassabis jokes Genie 4 launch means we live in simulations 'all the way down' — ZeroStateReflex · 2026-11-20
- Gemini 4 Neon places 4th in Sokoban eval as Hydrogen matches Terra at 7.5x cost — banteg · 2026-11-20
- Google confirms its 'most ambitious' Gemini 4 pre-training run is underway — OfficialLoganK · 2026-11-20
- LFM3-1.7B impresses in medical long-context work: 20k tok/s prefill on a 3090Ti — AdRepulsive7837 · 2026-11-20
- Google ships Gemini 4 Neon between Carbon and Hydrogen, 65.4% on CursorBench 4.1 — shensi · 2026-11-20
- MazeBench: GPT-6.2 Sol narrows gap with Astra but still scores less at same price — patience_cave · 2026-11-20