SnakeBench Average Apples
Games · 2025-02-04
SnakeBench average apples is the mean number of apples collected by the model across reported Snake matches. Higher values indicate better typical game performance.
Top models (higher is better)
| Model | Score |
|---|---|
| GPT-5.3-Codex | 16.8 |
| GPT-5.5 | 16.5 |
| GPT-5 | 15.7 |
| Opus 4.8 | 15.1 |
| o3 | 14.6 |
| Gemini 3.1 Pro Preview | 14.4 |
| Opus 4.7 | 14.4 |
| GPT-5.1 | 13.9 |
| GPT-5-Codex | 13.7 |
| Qwen3.5-Flash | 13.0 |
| Qwen3.6 35B A3B | 12.6 |
| GPT-5.1-Codex-Max | 12.3 |
| O4 Mini | 12.2 |
| Gemini 3.5 Flash | 11.9 |
| Qwen3.7-Max | 11.9 |
| Opus 4.6 | 11.7 |
| Gemini 2.5 Flash | 11.6 |
| Gemini 3 Pro Preview | 11.6 |
| GLM 5V Turbo | 11.5 |
| Devstral 2 | 11.5 |
| Gemini 3 Flash Preview | 11.2 |
| Doubao Seed 2.0 Mini | 11.2 |
| GLM-5.1 | 11.1 |
| GPT-5.2 | 11.1 |
| Qwen3.6 Plus (2026-04-02) | 11.0 |
| Gemini 2.5 Pro | 10.7 |
| Gemma 4 26B A4B | 10.6 |
| Sonnet 4.6 | 10.6 |
| Qwen3.5-27B | 10.6 |
| GPT-5.4 | 10.5 |
| Qwen3.5 122B A10B | 10.5 |
| GLM-5 | 10.4 |
| GPT-5.1-Codex mini | 10.2 |
| MiniMax M2.7 | 10.2 |
| o3-mini | 10.2 |
| Qwen3.5 35B A3B | 9.9 |
| Gemini 3.1 Flash-Lite | 9.9 |
| Doubao Seed 1.6 | 9.8 |
| Claude 3.7 Sonnet | 9.3 |
| Gemini 2.5 Flash Preview (09-2025) | 9.3 |