Overview
MiMo-V2-Omni is the omnimodal base model Xiaomi released in the first half of 2026, and Xiaomi's first generation to jointly model images, video, audio, and text within a single architecture. Cross-modal reasoning was designed as a default capability in this generation rather than a set of encoders bolted onto a text model—watching a video and answering a spoken question, or reading an image and producing audio commentary, were the differentiators it initially established.
One important note: the MiMo-V2 series was officially retired at 00:00 on June 30, 2026, and the original model names no longer work. Before that, Xiaomi had already routed V2 series requests to the MiMo-V2.5 series and billed at V2.5 pricing (V2-TTS switched earlier, from 00:00 on June 27). This page preserves the product record for the V2-Omni generation; the current in-service omnimodal mainstay is MiMo-V2.5-Omni.
Its contemporaries were MiMo-V2-Pro (reasoning/coding flagship) and MiMo-V2-Flash, and together the three supported Xiaomi's first-generation end-cloud AI experience across phones, cars, and smart home devices.
Key Features
- Native Omnimodal Unified Architecture: Any input and any output across images/video/audio/text, with cross-modal reasoning built into the architecture
- Video and Long Audio Understanding: Native support for understanding, summarizing, and Q&A on tens-of-minutes-long videos and audio, the headline strength of this Omni generation
- Interleaved Image-Text Reasoning: Handles interleaved input of "text + image + text + image" for complex reasoning, covering papers, technical documents, and multi-image manuals
- Open Source under MIT License: Weights and inference code are open under the MIT license, allowing commercial use and fine-tuning
- End-Cloud Synergy Deployment: A distilled version can run on Xiaomi HyperOS on-device, while the cloud version provides full capabilities, ensuring a consistent experience across Xiaomi's product line
Use Cases
- Full-modal assistant capabilities on Xiaomi phones/cars/smart home devices
- Video content understanding: automatic summarization, chapter segmentation, Q&A, keyframe localization
- Education scenarios: interleaved reasoning with images, text, and formulas, subject Q&A, video lecture explanations
- Industry/manufacturing: defect recognition directly from camera video or images, plus text report generation
- Accessibility: image description, video commentary, cross-modal dialogue, benefiting visually/hearing-impaired users
- Developers fine-tuning industry-specific full-modal models based on open-source weights
Pros
- First-generation native omnimodal architecture that set the technical direction for Xiaomi's later multimodal lineup
- Video and long audio understanding were this generation's main strengths
- MIT license open source, with weights open for commercial use and fine-tuning
- Paired with V2-Pro and V2-Flash in the same generation, covering reasoning, coding, and omnimodal scenarios
- End-cloud synergy: on-device distillation plus full cloud version, amplifying value across Xiaomi's entire product matrix
Pricing
Open-source weights are free (MIT license, downloadable from HuggingFace/GitHub). Official API (mimo.xiaomi.com) offers a free trial quota, with overage billed per token. Xiaomi phone/car users can use full-modal capabilities completely free in system-integrated scenarios. Enterprise private deployment can be quoted per GPU node.
Note: The MiMo-V2 series was retired at 00:00 on 2026-06-30 and the original model names no longer work. Please switch to the MiMo-V2.5 series and settle at V2.5 pricing.
Summary
MiMo-V2-Omni was the first generation of Xiaomi's omnimodal line, placing image, video, audio, and text modalities into a single architecture and open-sourcing it under MIT—a fairly sincere positioning among domestic open-source omnimodal releases at the time. The series was retired on 2026-06-30 and the model names no longer work; the current in-service omnimodal mainstay is MiMo-V2.5-Omni (1M context / 128K output / dual-protocol endpoints). For reasoning and coding use MiMo-V2.5-Pro; for speech synthesis and recognition use MiMo-V2.5-TTS and MiMo-V2.5-ASR respectively.
Version History
- MiMo-V2 Series Officially Retired, Omnimodal Role Taken Over by V2.5-Omni (2026-06-30): MiMo-V2-Pro / V2-Omni / V2-Flash were officially retired at 00:00 Beijing time on June 30, 2026, and the original model names stopped working. V2 series requests had already been routed to the V2.5 series and billed at V2.5 pricing; V2-TTS switched even earlier, auto-forwarding to MiMo-V2.5-TTS from 00:00 on June 27, 2026. The omnimodal mainstay thereby moved to MiMo-V2.5-Omni.
- MiMo-V2-Omni Released (2026-03-18): Xiaomi released its first-generation native omnimodal base model MiMo-V2-Omni, with any-input/any-output across images, video, audio, and text, cross-modal reasoning built into the architecture, MIT-licensed open weights, forming the first-generation MiMo product matrix alongside MiMo-V2-Pro and MiMo-V2-Flash.