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

148 models in hard prompts ranked by LMArena score · updated Nov 11, 2026

#ModelScore
148Qwen1.5-4B-ChatQwen976
147gemma-2b-itGoogle989
146OLMo-7B-InstructAi2993
145gemma-1.1-2b-itGoogle1005
144Llama-2-7b-chat-hfMeta Llama1009
143Qwen-14B-ChatQwen1027
142Phi-3-mini-128k-instructMicrosoft1028
141gemma-7b-itGoogle1042
140Llama-3.2-1B-InstructMeta Llama1044
139Llama-2-13b-chat-hfMeta Llama1051
138SOLAR-10.7B-Instruct-v1.0Upstage1062
137Qwen1.5-7B-ChatQwen1065
136Mistral-7B-Instruct-v0.2Mistral AI1067
135deepseek-llm-67b-chatDeepSeek1070
134gemma-1.1-7b-itGoogle1071
133Llama-2-70b-chat-hfMeta Llama1073
132granite-3.0-2b-instructIBM Granite1073
131Phi-3-mini-4k-instructMicrosoft1087
130granite-3.0-8b-instructIBM Granite1092
129Llama-3.2-3B-InstructMeta Llama1095
Show all 148 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.