Xiaomi releases MiMo‑V2‑Flash model with faster response than DeepSeek
Xiaomi has officially launched and open‑sourced its new foundational language model, MiMo‑V2‑Flash. The model emphasizes high efficiency and ultra‑fast performance, excelling in reasoning, code generation, and agent‑based applications. It can also serve as a general AI assistant for everyday tasks.
Early user testing shows that MiMo‑V2‑Flash delivers response speeds faster than other models such as Doubao, DeepSeek, and Yuanbao. Many users reported being surprised by the speed difference, especially in question‑answering scenarios.
From a technical standpoint, MiMo‑V2‑Flash has a total of 309 billion parameters, with 15 billion active parameters. Benchmark results place it among the top tier of current open‑source large models. Both the model weights and inference code are fully open‑sourced under the MIT license.
In terms of cost, Xiaomi has priced the API at $0.10 (≈¥0.73) per million input tokens and $0.30 (≈¥2.19) per million output tokens. For now, the company is offering limited free access to encourage adoption.
Xiaomi is expected to reveal more technical details during its full‑ecosystem partner conference, scheduled for today. This event may provide deeper insights into MiMo‑V2‑Flash's architecture and future applications.
Xiaomi, founded in 2010 and headquartered in Beijing, is one of China's largest consumer electronics and technology companies. Known globally for its smartphones, smart home devices, and IoT ecosystem, Xiaomi has expanded into electric vehicles and artificial intelligence research. The company has consistently invested in open‑source projects and advanced R&D, aiming to strengthen its position in both domestic and international technology markets.
