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

84 models in chinese ranked by LMArena score · updated Nov 11, 2026

#ModelScore
84Llama-3.2-1B-InstructMeta Llama959
83Llama-3.2-3B-InstructMeta Llama1017
82granite-3.0-8b-instructIBM Granite1063
81granite-3.0-2b-instructIBM Granite1070
80Mixtral-8x22B-Instruct-v0.1Mistral AI1116
79granite-3.1-2b-instructIBM Granite1139
78granite-3.1-8b-instructIBM Granite1145
77Llama-3.1-8B-InstructMeta Llama1151
76Llama-3.1-70B-InstructMeta Llama1215
75Llama-3.3-70B-InstructMeta Llama1217
74QwQ-32B-PreviewQwen1218
73Qwen2.5-Coder-32B-InstructQwen1222
72Llama-3.1-405B-InstructMeta Llama1242
71granite-4.0-h-smallIBM Granite1251
70Llama-4-Scout-17B-16E-InstructMeta Llama1255
69Qwen2.5-72B-InstructQwen1272
68Llama-4-Maverick-17B-128E-InstructMeta Llama1278
67granite-4.2-3bIBM Granite1279
66Olmo-3.1-32B-InstructAi21304
65gpt-oss-20bOpenAI1315
Show all 84 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.