Xiaomi releases MiMo-V2.5, an open multimodal agent model with 1M context
Original titleXiaomi MiMo-V2.5
AISummary
Xiaomi released MiMo-V2.5, a 310B-parameter sparse MoE model with 15B active parameters that adds native visual and audio understanding. The model supports up to 1 million tokens of context, and its weights, tokenizer, and model card are available on Hugging Face. Xiaomi says it surpasses MiMo-V2-Pro on agentic performance and reports a Claw-Eval score of 62.3 on the general subset.
AIWhy it matters
The release pairs native visual and audio understanding with a 1M-token context window and open weights, a combination worth checking against your own multimodal workflows.
Source: Xiaomi MiMo · mimo.xiaomi.comPublished · added here