Xiaomi has made a move that just a few years ago almost nobody expected. From a manufacturer of smartphones and gadgets, it’s increasingly becoming a significant player in AI as well.
The biggest Xiaomi community group
It has now made its own ChatGPT-like platform available to the public for the latest time, called Xiaomi MiMo Studio. You can try their newest model MiMo-V2-Flash directly in your browser. No app installation, no registration—just open the website and start chatting.
MiMo Studio: Xiaomi AI available to everyone
MiMo Studio is an online AI chat platform that Xiaomi launched as a testing environment for its latest language model. You’ll find a classic chat mode here, but also advanced options like deep logical reasoning or answers enhanced with online search. In practice, that means it’s useful not only for everyday questions, but also for programming, text analysis, research, or solving more complex tasks.
Xiaomi openly admits that demand has exceeded expectations. The servers are sometimes overloaded and you may need to sign in multiple times, but the platform is available globally, without regional restrictions.
For example, at the time of writing this article, the AI chatbot website wasn’t working, but we believe it will be functional again soon and you’ll be able to try it too. You can find the new AI chatbot at this internet link: aistudio.xiaomimimo.com.
MiMo-V2-Flash: speed, understanding, and huge numbers
The whole project is powered by the MiMo-V2-Flash model, which Xiaomi developed in-house. It’s a Mixture-of-Experts (MoE) model designed specifically for intelligent agents. Overall, it has up to 309 billion parameters, and during processing it actively uses about 15 billion. That makes it significantly more efficient than traditional huge models.
The architecture combines global attention with Sliding Window Attention at a ratio of 1:5, enabling it to work with long context. It natively handles 32k tokens, but it’s scalable up to 256k. That’s a level only a narrow top tier of today’s AI models can manage.
According to internal benchmarks, Xiaomi places MiMo-V2-Flash among the two best open-source models in the world in the area of agent tasks. Its strongest point should be logic, programming, and multi-step reasoning. Xiaomi goes even further and claims that in these areas its performance approaches that of closed commercial models, which typically cost several times more.
The economic aspect is also interesting. Xiaomi talks about reducing inference costs to about 2.5% of the price of comparable closed models, while answer generation speed is supposed to be roughly twice as fast. That makes MiMo-V2-Flash a very attractive solution for businesses and developers.
Open-source, MIT license, and cheap API
Another thing that Xiaomi nailed is openness. The model and the inference code are released under the MIT license, so you can use, modify, and integrate them freely into your own projects. For developers, there’s also an API with prices that are significantly lower than those of big competitors.
Xiaomi charges about $0.10 per million input tokens and $0.30 per million output tokens. Plus, during a limited time, the API is available for free, making it much easier to test and experiment.
To get more updates from the Xiaomi world, keep following our Xiaomi Planet website — join our Facebook group, where we build an active community and help each other out. Also subscribe to our Telegram, where we add the best offers and coupons from China every day.
Don’t miss the best coupons 📩
Subscribe and we’ll email you new coupons before they expire. We only send it once a month — no spam.
