Xiaomi MiMo
309B total parameters with 15B active: Xiaomi MiMo uses MoE to push cost down while stretching context to 1M. A 7B reasoning series started the line, V2-Flash, V2-Pro, Omni and TTS extend it into a full agent stack, and V2.5 ships open under MIT.
MiMo started as a 7B reasoning model, then moved to MoE: V2-Flash carries 309B total parameters with 15B active and 256K context for multi-turn tool use, while V2-Pro goes past 1T total and 42B active at 1M context, opening its API first, with Omni on multimodal understanding and TTS on speech. V2.5 ships under MIT at 1.02T/42B, reported near the top of GDPVal-AA and ClawEval. Deployment guides recommend SGLang or vLLM with FP8.
