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

147 models in multi-turn ranked by LMArena score · updated Nov 11, 2026

#ModelScore
147gemma-2b-itGoogle945
146gemma-1.1-2b-itGoogle958
145gemma-7b-itGoogle963
144Qwen1.5-4B-ChatQwen977
143Phi-3-mini-128k-instructMicrosoft989
142OLMo-7B-InstructAi21007
141Qwen-14B-ChatQwen1022
140Llama-2-7b-chat-hfMeta Llama1029
139Llama-3.2-1B-InstructMeta Llama1030
138gemma-1.1-7b-itGoogle1040
137SOLAR-10.7B-Instruct-v1.0Upstage1045
136Phi-3-mini-4k-instructMicrosoft1049
135Llama-2-13b-chat-hfMeta Llama1050
134granite-3.0-2b-instructIBM Granite1053
133Qwen1.5-7B-ChatQwen1061
132Mistral-7B-Instruct-v0.2Mistral AI1062
131granite-3.0-8b-instructIBM Granite1063
130Phi-3-small-8k-instructMicrosoft1068
129deepseek-llm-67b-chatDeepSeek1082
128Phi-3-medium-4k-instructMicrosoft1088
Show all 147 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.