With You, Every Step of the Way
Apply for MiMo-V2.5-Pro-UltraSpeed Early Access
Full-power performance at 1,000 tokens/s peak. Unlocks the full potential of Coding Agents. Limited slots — daily quota, priority access for qualified organizations.
MiMo Claw Official Launch — Limited-Time Offer ¥14.9/mo
Flagship mimo-v2.5-pro × native OpenClaw integration × WPS ecosystem — three major upgrades, stackable with Token Plan, free usage extended 1h → 4h/day.
Xiaomi MiMo-V2.5 Series — Now Available
Sharper instruction following, better omni-modal understanding, more natural speech synthesis.
Xiaomi MiMo Token Plan
Simple, transparent, and great value. Monthly or annual plans covering the full V2.5 model series.
MiMo-V2.5-Pro
Trillion-Scale, Efficient Architecture: 1T total params | 42B active | 1M context. Peak Agent Performance: Rivals Claude Opus 4.6 in demanding agentic workloads.
Input (cache hit)$0.0036 / MTok
Input (cache miss)$0.435 / MTok
Output$0.87 / MTok
MiMo-V2.5-Pro-UltraSpeed
MiMo Algorithm Innovation：Combines FP4 lossless quantization and DFlash parallel decoding to maximize inference throughput. TileRT System-level Optimization：Optimizes GPU efficiency using persistent kernels for compute-transfer overlap and heterogeneous pipelines.
Input (cache hit)$0.0108 / MTok
Input (cache miss)$1.305 / MTok
Output$2.61 / MTok
MiMo-V2.5
Native Omni-Modal Understanding + 1M Context: Natively processes images, video, audio, and text for cross-modal perception and long-range reasoning. Powerful Omni-Modal Agent Capabilities: Native agent execution for complex tasks: browse, comprehend, reason, and act.
Input (cache hit)$0.0028 / MTok
Input (cache miss)$0.14 / MTok
Output$0.28 / MTok
MiMo-V2.5-TTS Series
Premium Voice TTS: Built-in high-quality voices with fine-grained control over pace, emotion, and tone. Voice Design & Cloning: Create a new voice from a single sentence, or clone any voice with just a few audio samples.
MiMo-V2.5-ASR
Bilingual + Dialect, Lyrics Transcription: Supports Chinese and English recognition across regional dialects; transcribes lyrics from mixed vocal and instrumental audio. Robust in Challenging Acoustics: Reliable in high-noise, far-field, and multi-speaker conditions; accurate on knowledge-dense content.
Per audio hour$0.074 / hour
Xiaomi MiMo Code
Next-gen AI coding assistant for developers: Empowering you to understand, build, and collaborate faster with infinite context.
Xiaomi MiMo Claw
An all-in-one agentic platform for real-world tasks. Plan, call tools, reason across steps, and execute autonomously — AI that goes beyond chat.
Xiaomi MiMo Studio
Chat with MiMo, no setup needed. Switch between model versions, explore multimodal interactions and long-context reasoning — all in one place.
Xiaomi MiMo API
The developer platform for MiMo model access. OpenAI- and Anthropic-compatible APIs, comprehensive docs, low-latency inference — integrate MiMo into your app in minutes.
We distill patterns from vast data, compressing complexity into language and action you can understand. We believe intelligence is rooted in empathy — its true value lies in human connection.
MiMo is more than a tool behind a screen — it wants to be your real-world companion: reading your words, seeing your world, sparking your ideas, and exploring the frontiers of intelligence together.