Meituan unveils LongCat-2.0, claims breakthrough AI Model trained entirely on Chinese chips
Chinese technology company Meituan has announced the release of LongCat-2.0, a next-generation large language model (LLM) that the company says was developed entirely using Chinese-made processors, marking a significant milestone in the country's drive for greater technological self-reliance.
The Beijing-based food delivery and local services giant said Tuesday that the new model will be made open source, allowing developers and researchers to access and build upon the technology.
According to Meituan, LongCat-2.0 is the world's first trillion-parameter AI model both trained and operated on a computing cluster consisting of approximately 50,000 domestically manufactured chips. If confirmed, the achievement would represent a major advance for China's AI ecosystem, which has long depended on high-performance processors designed by U.S. companies.
The announcement comes as Chinese technology firms continue to accelerate investments in artificial intelligence, seeking new growth opportunities beyond their traditional businesses. Meituan joins a growing list of companies expanding into generative AI, competing with domestic players such as DeepSeek and ByteDance's Doubao.
Unlike many of its rivals, Meituan entered the AI race relatively recently. The company's LongCat research team was established in 2023, while its first large language model was introduced only late last year.
The launch also reflects the broader impact of U.S. export restrictions on advanced semiconductor technology. Washington's controls have limited Chinese companies' access to cutting-edge AI chips, prompting major technology firms to accelerate the development and deployment of domestic alternatives.
By relying exclusively on Chinese processors, Meituan is positioning LongCat-2.0 as evidence that advanced AI systems can be developed using homegrown computing infrastructure, reducing dependence on foreign semiconductor technologies.
The company has not disclosed detailed performance benchmarks comparing LongCat-2.0 with leading international or domestic AI models. However, the decision to release the model as open source is expected to encourage broader adoption and collaboration within China's rapidly evolving AI community.
-
19:34
-
19:19
-
18:58
-
18:41
-
18:21
-
18:05
-
17:45
-
17:27
-
17:12
-
16:48
-
16:32
-
16:16
-
16:00
-
15:42
-
15:24
-
15:10
-
14:45
-
14:26
-
14:09
-
13:45
-
13:25
-
13:10
-
12:47
-
12:33
-
12:16
-
12:00
-
11:40
-
11:25
-
11:11
-
11:00
-
10:49
-
10:34
-
10:17
-
10:12
-
09:58
-
09:41
-
09:20
-
09:05
-
08:48
-
08:35
-
08:18
-
07:57
-
07:41
-
07:19
-
22:00
-
21:30
-
21:10
-
20:48
-
20:32
-
20:15
-
19:50
