BEIJING: Meituan on Tuesday unveiled and open-sourced its latest large language model, saying it was trained entirely on Chinese-made chips, marking a significant step in the country’s drive for artificial intelligence self-reliance amid US export restrictions.
The company said its new LongCat-2.0 model is the world’s first trillion-parameter AI system trained exclusively on a cluster of 50,000 domestically produced chips. Meituan said the model is designed for tasks including coding, website development and processing long documents of up to one million tokens.
Why Does This Matter?
The launch highlights China’s efforts to reduce dependence on foreign semiconductor technology as US export controls continue to restrict Chinese companies’ access to advanced AI chips.
Chinese firms have increasingly turned to domestically developed processors to train advanced AI models, with companies such as Huawei and Enflame expanding their presence in the sector.
How Will Meituan Use the AI Model?
Meituan said LongCat-2.0 will be integrated into its digital services, including restaurant recommendations, hotel bookings and other consumer applications, as the company expands its AI capabilities.
The company added that the model has performed competitively against leading AI systems, including OpenAI’s GPT-5.5 and Google’s Gemini, according to its internal evaluations.
What Does It Mean for China’s AI Industry?
The release underscores intensifying competition in China’s AI sector, where technology companies are investing heavily in developing advanced models using domestic computing infrastructure.
The move also reflects Beijing’s broader strategy to strengthen technological self-sufficiency and accelerate AI development despite ongoing restrictions on access to high-end foreign chips.