Atlas

Benchmarks

← All benchmarks

SnakeBench Total Apples

Games · 2025-02-04

SnakeBench total apples counts how many apples the model collected across the match set reported on its model page. Higher values indicate stronger gameplay outcomes.

Top models (higher is better)

ModelScore
MiniMax M31294
Kimi K2.51196
GPT-5864
GPT-5.3-Codex772
GPT-5.1-Codex-Max392
Gemini 3.1 Pro Preview342
GPT-5.1320
GPT-5.1-Codex mini317
Qwen3 VL 235B A22B Thinking301
Qwen3 32B277
Gemini 3 Pro Preview267
GPT-5 Chat (2025-08-07)251
Qwen3-235B-A22B242
GPT-5 Nano237
Qwen3-Next-80B-A3B-Thinking234
Gemini 2.5 Flash221
Qwen3.5-Flash221
Gemini 3 Flash Preview213
Opus 4.8211
GPT-5-Codex205
o3205
Qwen-Plus-2025-07-28203
Grok Code Fast 1192
Sonnet 4.6191
GLM-4.7191
Qwen3-Coder-Plus (rolling alias)187
GPT-5.5181
Opus 4.6176
Olmo 3 32B Think174
Kimi K2 Thinking170
Doubao Seed 2.0 Mini168
Qwen3 14B166
Sonnet 4.5163
DeepSeek-V3.1-Terminus162
Grok 4 Fast160
Qwen3.5-27B159
Opus 4.7158
Llama 4 Maverick Instruct158
Qwen3.6 Plus (2026-04-02)154
Claude 3.7 Sonnet154
Loading Atlas data…