Xiaomi launches its first MiMo AI model with 7 billion parameters
Chinese smartphone and electric car maker Xiaomi says its open-source MiMo reasoning model, fully trained in-house, rivals OpenAI’s o1-mini and Alibaba’s QwQ-32B in performance
Chinese smartphone and electric car maker Xiaomi says its open-source MiMo reasoning model, fully trained in-house, rivals OpenAI’s o1-mini and Alibaba’s QwQ-32B in performance
This open-source model includes 7 billion different parameters. MiMo is said to have outperformed OpenAI’s o1-mini and Alibaba Group’s QwQ-32B-Preview in performance tests.
According to the South China Morning Post, MiMo is Xiaomi’s first large-scale language model (LLM), and the company said it was developed using reinforcement learning by its specialized AI working group, known as Core.
Xiaomi’s share price in Hong Kong rose 5,3% to HK$49.95 on Friday, while shares of Kingsoft Cloud Holdings, in which Xiaomi owns a 10% stake and Xiaomi CEO Lei Jun owns 11%, jumped 14,2% to HK$7.4.
The model’s launch is consistent with previous reports that Xiaomi is ramping up its computing resources. According to a report by local media outlet Jiemian in December, Xiaomi has purchased about 10,000 GPUs to train its models. Xiaomi’s AI ambitions became apparent when the company offered to hire Luo Fuli, the «genius girl» of Chinese AI at DeepSeek. Luo, a key contributor to the development of the DeepSeek-V2 model, ultimately declined the offer.
Last May, Xiaomi announced plans to develop its own LLM as it seeks to incorporate its own AI technology into its electric vehicles, smartphones and home appliances.
Although Xiaomi is releasing a reasoning model following DeepSeek and Alibaba, the company is confident that they still have a chance to reach the theoretical point where AI equals or surpasses human intelligence.
OpenAI aims to teach its «open» AI model to ask for help from cloud models
OpenAI aims to teach its «open» AI model to ask for help from cloud models
Alibaba launches Qwen3, a family of «hybrid» AI reasoning models
Alibaba launches Qwen3, a family of «hybrid» AI reasoning models
Read the country’s main IT news in our Telegram
Read the country’s main IT news in our Telegram
