From RedNote to Meituan, digital platforms are bypassing third-party providers to build in-house AI models
While China’s new “AI tigers” have captured global attention for challenging Silicon Valley front-runners, a quieter wave of domestic technology companies is building in-house artificial intelligence systems serving everyday users.
Beyond pure-play AI labs like DeepSeek, Z.ai (also known as Zhipu AI), Moonshot AI and MiniMax, consumer-facing platforms that span e-commerce, video gaming, social media and travel are developing foundation models tailored to their own business ecosystems, in efforts that have drawn far less attention.
RedNote, the lifestyle platform often described as China’s answer to Instagram, underscored the trend earlier this month with its latest release.
Its AI research arm, Dots Studio, introduced Dots3-Note Preview, a 280-billion-parameter open-weight model. Citing benchmark tests, the company said the system matched or outperformed models from US labs OpenAI and Anthropic, as well as domestic leaders DeepSeek and Z.ai, in specific tasks.
The release highlights a growing technical focus for a platform best known for fashion, travel and shopping recommendations. Dots Studio said in a blog post that it aimed to “build AI that benefits everyone and helps people solve the many problems they encounter in life”.
Industry analysts said the effort reflected a broader operational shift across China’s technology sector, where digital platforms are building custom AI rather than relying solely on third-party providers.