Xiaomi has unveiled the new Xiaomi AI Model, MiMo‑V2‑Flash, as a significant move into the generative AI ecosystem. This open‑source, innovative system aims to compete with top-tier AI models and deliver strong capabilities for developers and creators worldwide.
What is Mimo V2 Flash?
MiMo-V2-Flash is an advanced open-source Xiaomi AI model with 309 billion parameters, which uses 15 billion active per task. It is designed to handle reasoning, coding, agentic, and general-purpose tasks in an efficient manner.
Performance and benchmarks
Xiaomi stated that MiMo‑V2‑Flash can achieve inference speeds of up to 150 tokens per second and has been compared to other leading AI models. In reasoning tests, it can be on par with DeepSeek V3.2 and Moonshot AI’s Kimi K2. In contrast, on agentic tasks, it achieved a high score on the SWE‑Bench Verified, outperforming many open-source peers and approaching the performance of OpenAI’s GPT‑5‑High.
Accessibility and open‑source approach
This Xiaomi AI model acts as a lighthouse point because of its open-source character. MiMo‑V2‑Flash is accessible across platforms, including Xiaomi’s developer portal, MiMo Studio, Hugging Face, and Xiaomi’s API services. When it is put out under an open license, it practically means that developers are at liberty to test, change, and create by the model, which may lead to a fast innovation turn in the global AI community.
Cost-effective and scalable
MiMo‑V2‑Flash could differ significantly from other devices due to its low price. According to Xiaomi, inference costs are only $0.1 per million input tokens and $0.3 per million output tokens. This means that advanced generative AI becomes less costly and thus can be used by startups, enterprises, and individual developers without any barrier to entry.
FAQ(Frequently Asked Questions)
Q1: What is the Xiaomi AI Model MiMo‑V2‑Flash?
MiMo‑V2‑Flash is an open-source generative AI model by Xiaomi, capable of reasoning, coding, agentic tasks, and general-purpose use, with a Mixture-of-Experts architecture comprising 309 billion total parameters.
Q2: How can developers access MiMo‑V2‑Flash?
Xiaomi’s developer portal, MiMo Studio, Hugging Face, and Xiaomi’s API services provide access to the model. Developers are thus free to experiment, modify, and create applications
Q3: What makes MiMo‑V2‑Flash cost-effective and scalable?
MiMo‑V2‑Flash is characterized by very low inference costs $0.1 per million input tokens and $0.3 per million output tokens thus a cutting-edge generative AI technology is made affordable to startups, enterprises, and individual developers.
Xiaomi unveils open-source AI model Mimo V2 Flash to compete in generative AI ecosystem