Through an API

Open and closed models answer the same suite-standard 1.0 through their APIs, ranked together. 1 ranked model.

#ModelPass rate
1DeepSeek V4.1 Flash 1017Open weights100%

Model leaderboard

81 models in math ranked by LMArena score · updated Nov 11, 2026

#ModelScore
81Llama-3.2-1B-InstructMeta Llama1086
80granite-3.0-2b-instructIBM Granite1117
79Llama-3.2-3B-InstructMeta Llama1126
78granite-3.0-8b-instructIBM Granite1143
77granite-3.1-8b-instructIBM Granite1152
76granite-3.1-2b-instructIBM Granite1159
75Llama-3.1-8B-InstructMeta Llama1179
74Mixtral-8x22B-Instruct-v0.1Mistral AI1184
73QwQ-32B-PreviewQwen1213
72Qwen2.5-Coder-32B-InstructQwen1251
71Llama-3.1-70B-InstructMeta Llama1252
70granite-4.0-h-smallIBM Granite1253
69Llama-3.3-70B-InstructMeta Llama1267
68Llama-3.1-405B-InstructMeta Llama1281
67Qwen2.5-72B-InstructQwen1283
66Llama-4-Scout-17B-16E-InstructMeta Llama1287
65Llama-4-Maverick-17B-128E-InstructMeta Llama1300
64Olmo-3.1-32B-InstructAi21307
63granite-4.1-8bIBM Granite1313
62gpt-oss-20bOpenAI1317
Show all 81 models

Score by size

Plot against
Best for its size

Scores: LMArena, CC BY 4.0. Measured on served versions, not local builds. A model with several served variants shows its best.