Xiaomi Joins The Chinese AI Race With The Release Of MiMo: A Powerful Open-Source AI Model for Reasoning and Code Tasks

Chinese smartphone and electric vehicle maker, Xiaomi has officially stepped into the AI race with the launch of MiMo, its first open-source large language model (LLM) engineered specifically for advanced reasoning tasks.

Despite its relatively compact architecture of 7 billion parameters, MiMo is already outperforming larger competitors like OpenAI's o1-mini in math reasoning and coding, positioning Xiaomi as a serious contender in the AI space.

MiMo is Xiaomi's first language model, and the company said it was developed using reinforcement learning by its specialised AI task force, known as Core.

Xiaomi emphasized that MiMo is particularly well-suited for tasks that demand logical thinking, mathematical reasoning, and complex inference. This makes it a powerful tool for domains like algorithm design, software engineering, data science, and academic research.

Its smaller size also offers a unique advantage: MiMo can be deployed in enterprise environments or edge computing setups where hardware resources are limited, making it both powerful and practical.

In line with the open-source philosophy, Xiaomi has made MiMo and its full technical report publicly accessible on GitHub and Hugging Face. This move invites the global developer community to test, fine-tune, and expand the model’s capabilities—potentially accelerating innovation and adoption.

Xiaomi's AI model comes at a time when China's deep-pocketed Big Tech firms are beginning to show their strength in developing foundational models, driven by the commercial potential of combining their products with AI.

While Xiaomi is releasing a reasoning model after DeepSeek and Alibaba, it said on Friday that it still has a chance to reach artificial general intelligence, the theoretical point at which AI equals or surpasses human intelligence.

"The year 2025 seemed to be the second half of the AI model competition, but we firmly believe the road to AGI is still very long."

Alibaba has officially rolled out the third generation of its open-source Qwen AI models, aiming to raise the bar in speed, scalability, and multilingual processing.

The launch adds fresh momentum to China’s already competitive AI landscape, which has seen a rapid influx of powerful language models from tech giants across the country.

The new Qwen3 series spans eight model sizes—from a lightweight 600 million parameters to a massive 235 billion—each boasting upgrades in efficiency, reasoning, and language versatility.

According to Alibaba Cloud, the division leading the company’s AI development, all versions of Qwen3 have been optimized to deliver faster processing and broader language support, making them more accessible for both enterprise and developer use.

Other major players like ByteDance and Tencent Holdings have also made significant strides in AI. Just last week, Baidu announced the launch of Ernie 4.5 Turbo, its latest iteration of large language models, further intensifying competition among China's leading tech firms.