Meituan says it has trained a new large AI model, LongCat-2.0, end-to-end using home-grown Chinese silicon. According to the company, the model has 1.6 trillion parameters and is positioned as the first model of its size to be trained entirely on domestic chips. The Next Web reports that Meituan highlights this training approach in response to U.S. export controls, framing the effort as a way to reduce reliance on restricted hardware.
Investing.com and The Next Web both report that Meituan’s announcement centers on the use of domestic chips for training the new model. While the sources describe the scale of the model and the claim about end-to-end domestic training, they do not provide additional technical specifications or independent verification details. Overall, the coverage indicates that Meituan is emphasizing local AI compute capabilities as a key feature of LongCat-2.0.
