MiMo-V2

Focuses on the MiMo-V2 line from Xiaomi and compares Flash, Pro, Omni, TTS and V2.5 across parameter scale, context length and multimodal capability.

MiMo-V2 is the second-generation model line from Xiaomi MiMo, spanning Flash, Pro, Omni and TTS branches. The open V2-Flash keeps inference cost down at a 256K context through a sparse 309B-parameter design that activates 15B, while V2-Pro and V2.5 widen the window to 1M and strengthen complex task orchestration, code engineering and multimodal understanding.

Xiaomi MiMo-V2.5 Open Source Interpretation: MIT Protocol, 1M Context, and Native Multimodal Capabilities

1. Abstract The Xiaomi MiMo-V2.5 series has been officially open source, using the MIT License, and the weights and model cards are published on Huggi...

Admin 4/28/2026

484

Xiaomi released MiMo-V2-Pro, Omni and TTS, betting on the Agent full-stack model

The Xiaomi MiMo team launched MiMo-V2-Pro, MiMo-V2-Omni and MiMo-V2-TTS, positioning it as a full-stack model family for the Agent era. Existing offic...

Admin 3/19/2026

216

MiMo-V2-Flash Open Source Interpretation: 309B MoE, 15B Activation Parameters and 256K Long Context

1. Abstract MiMo-V2-Flash is an open-sourced hybrid expert (MoE) large language model from the Xiaomi MiMo team, with a total parameter of about 309B ...

Admin 12/17/2025

463

MiMo-V2-Flash released: 256K long context and multi-token prediction to improve inference throughput

Xiaomi MiMo and the Xiaomi large model Core team have released and opened MiMo-V2-Flash-related resources, positioning it as a basic language model fo...

Admin 12/17/2025

229