Skip to content
snakebench

LLM benchmark

Plan every move.
Before the first one.

SnakeBench gives a model the full map and every future food, then asks for the whole game as one string. No feedback. No second chances. Just planning.

Run the benchmark How it works
reference solver ยท seed 2026 food 0/20

The loop

  1. 01

    Generate

    Pick a grid, an obstacle count and a seed. Same seed, same board, every time.

  2. 02

    Reveal

    The model sees everything up front: walls, its body, and every food that will ever spawn.

  3. 03

    Plan

    It replies with one compressed string like 6L1R2L1L6. No retries, no feedback.

  4. 04

    Simulate

    The plan runs deterministically. Score is food eaten before the first mistake.

Snake is easy to play.
It is hard to plan blind.

  • Long-horizon planningDozens of foods, hundreds of moves, one answer.
  • Spatial reasoningCoordinates, walls and turns relative to facing.
  • State trackingThe body grows and trails behind. Never bite it.
  • OptimizationFewer moves breaks ties. Detours cost points.