Xiaomi has recently unveiled MiMo, its in-house open source AI model with reasoning capabilities. With the release of the large language model, the company has joined the ranks of China’s tech firms, hoping to gain a foothold in the AI industry.
According to Xiaomi, MiMo was developed by Core, a specialised AI task force. The smartphone company also claims that the AI model has seven billion parameters and has outperformed models like OpenAI’s o1-mini and Alibaba’s QwQ-32B-Preview in mathematical reasoning and coding benchmarks.
The South China Morning Post reported that Xiaomi had been focusing on developing its computing resources by purchasing roughly 10,000 graphics processing units to train its models. The company has also attempted to hire a DeepSeek employee who contributed to the DeepSeek-V2 model, although the offer was declined. It’s quite likely that the launch of MiMo is the culmination of such efforts.
Xiaomi’s foray into AI models is hardly a surprise given the increasing prevalence of AI in commercial products like smartphones. It’s quite likely the company wishes to integrate its hardware products with AI that it has developed in-house. Of course, the emergence of DeepSeek’s R1 model has also fueled rapid AI development in China, so Xiaomi may also be looking to stay ahead of the technological curve. Alibaba has also recently revealed its latest generation of its Qwen models, further intensifying the competition in the AI field.
(Source: South China Morning Post)
