Xiaomi MiMo-V2.5-Pro achieves multiple breakthroughs on the Arena leaderboard
Xiaomi MiMo announced on April 26 that MiMo-V2.5-Pro has achieved multiple breakthroughs on the latest Arena leaderboard: Text Arena (Expert) ranks 6th globally and 1st among open-source models, while also ranking 1st among Chinese models; Xiaomi ranks 3rd globally by lab, behind only Anthropic and OpenAI. Text Arena (Overall) ranks 2nd among open-source models globally, Code Arena (WebDev) ranks 3rd among open-source models globally, and it ranks 1st among open-source models globally in four sub-leaderboards: Hard Prompts, Hard Prompts (English), Instruction Following, and Long Query. On April 28, Xiaomi MiMo further open-sourced the MiMo-V2.5 series under the MIT license, supporting commercial deployment, continued training, and fine-tuning without additional authorization; both models support a 1M token context window, with MiMo-V2.5-Pro targeting complex agent and coding tasks and ranking first among open-source models on GDPVal-AA and ClawEval, while MiMo-V2.5 is a native omni-modal model. SGLang and vLLM provided support on day one and completed Day-0 adaptation for multiple chip platforms.
Xiaomi MiMo announced on April 26 that MiMo-V2.5-Pro has achieved multiple breakthroughs on the latest Arena leaderboard: Text Arena (Expert) ranks 6th globally and 1st among open-source models, while also ranking 1st among Chinese models; Xiaomi ranks 3rd globally by lab, behind only Anthropic and OpenAI. Text Arena (Overall) ranks 2nd among open-source models globally, Code Arena (WebDev) ranks 3rd among open-source models globally, and it ranks 1st among open-source models globally in four su
Matched fieldsSummary