MiMo V2: From Perception to Action
Agentic foundation models by Xiaomi — Pro for long-context reasoning with 1M tokens, Omni for unified multimodal perception.
Built to call tools and complete real-world tasks.
Ranked #8 globally on Artificial Analysis Intelligence Index
1T+ tokens processed during early access
Works with your favorite AI frameworks
What is MiMo V2
Agentic intelligence for real-world tasks — Pro supports long-context reasoning, while Omni unifies multimodal perception with action.
Long-Context Reasoning
Maintain intent across 1M token contexts for coherent multi-step planning and generation.
Unified Multimodal Perception
Natively process image, video, and audio inputs within a single unified architecture.
Tool Calling & Task Completion
Structured function calling for multi-step agentic workflows with reliable execution.
Agentic Execution
End-to-end task completion in real digital environments — code, browse, and operate.
Choose Your Model
Two models tuned for real workflows: long-context reasoning, unified multimodal understanding, and reliable tool-enabled execution.
Get Started in 3 Steps
From zero to agentic AI in minutes:
Get Your API Key
Visit the MiMo Platform or OpenRouter and generate your API credentials in seconds.
Integrate with Your Stack
Use OpenRouter or direct API access. Compatible with OpenClaw, Cline, KiloCode, and standard OpenAI-format SDKs.
Build & Deploy
Start building agentic workflows, coding assistants, or multimodal applications with production-grade reliability.
Key Capabilities of MiMo V2
From perception and planning to structured tool calls and end-to-end task results.
1M Token Context Window
MiMo V2 Pro processes entire codebases and long documents without losing coherence across extended sessions.
Trillion-Parameter Architecture
42B active parameters with hybrid attention (7:1 ratio) for frontier-level quality and efficient inference.
Multi-Token Prediction
Specialized prediction layer for faster output while preserving reasoning depth on agentic tasks.
Native Reasoning
Built-in chain-of-thought with transparent decision-making. MiMo V2 exposes its reasoning for easy debugging.
Framework Compatible
Drop-in replacement for existing workflows — OpenClaw, Cline, KiloCode, Blackbox, and any OpenAI-compatible SDK.
Multimodal Native
MiMo V2 Omni processes text, image, video, and audio in one architecture — no separate preprocessing needed.
MiMo V2 by the Numbers
Performance that speaks for itself.
#8 Global AI Intelligence Ranking
Global AI Intelligence Ranking
1T+ Total Parameters (Pro)
Total Parameters (Pro)
1M Context Window Tokens
Context Window Tokens
$1/M Starting Input Price
Starting Input Price
What Developers Say
Hear from developers and teams building with these models.
The 1M context window changes everything. I feed entire repositories and get coherent refactoring plans that actually work.
Alex Chen, AI Engineer
Alex Chen
AI Engineer
We switched our agentic pipeline to MiMo V2 Pro. Tool-calling reliability is noticeably better and costs dropped significantly.
Sarah Kim, CTO at AgentLab
Sarah Kim
CTO at AgentLab
The coding capabilities are genuinely impressive. Pro generates production-ready frontend code in a single query — cleaner than what I'd write manually.
Marcus Liu, Full-Stack Developer
Marcus Liu
Full-Stack Developer
The unified multimodal architecture in Omni is elegant. Processing video, audio, and text in one model eliminates our entire preprocessing pipeline.
Priya Patel, ML Researcher
Priya Patel
ML Researcher
At $1/M input tokens, this is the best value frontier model. I run it at scale without worrying about API costs.
James Park, Indie Developer
James Park
Indie Developer
We integrated MiMo V2 into our customer support agent via OpenClaw. Resolution time dropped 40% compared to our previous provider.
Elena Volkov, Product Lead
Elena Volkov
Product Lead
Frequently Asked Questions
For updates and API access, visit the <a href='https://mimo.xiaomi.com' target='_blank' class='underline'>MiMo Platform</a>.
Start Building with MiMo V2
From perception to action — explore Pro & Omni on the official MiMo Platform.