Meituan Open Sources LongCat-2.0 Large Model, a Milestone for Domestic Computing Power

On July 6th, the AI field saw a major milestone as Meituan officially announced the open-source release of its foundational large model, LongCat-2.0. Industry experts regard this move as a significant breakthrough in applying domestic computing power to large-scale AI models.
As the first foundational large model in the industry trained and inferred entirely on domestic computing chips, LongCat-2.0 demonstrates impressive technical strength. Officially released on June 30th, both training and inference were successfully completed on a domestic computing cluster consisting of 50,000 cards, fully showcasing the stability and high performance of domestic chips when handling trillion-parameter-level models.
To promote technological inclusivity and ecosystem growth, Meituan adopted a comprehensive open-source strategy. On July 6th, the model weights, inference engine, and detailed technical documentation for LongCat-2.0 were simultaneously released on major developer platforms such as GitHub, Hugging Face, and ModelScope. Notably, the model is distributed under the MIT open-source license, allowing developers and companies to use it for commercial purposes free of charge, significantly lowering the barrier for small and medium-sized enterprises to adopt cutting-edge AI technology.
The domestic chip ecosystem responded quickly. On the same day Meituan announced the open-source release, three domestic chip manufacturers—Huawei Ascend, MoLeLineage, and Muxi Technologies—jointly announced that they had completed inference adaptation for the LongCat-2.0 model. This collaborative effort not only validates the model's compatibility with domestic hardware but also marks the rapid maturation of the integrated hardware-software ecosystem combining domestic chips with domestic large models.
This open-source initiative is not only a concentrated demonstration of technical capability but also provides a solid technical foundation for China's AI industry's self-reliance and control through the construction of a complete open-source ecosystem. As more developers contribute to optimizing and applying LongCat-2.0, the voice of domestic AI computing power in international competition is expected to grow further.
Related article
DeepMind Trio Behind Poker AI Now Generating Returns for Quant Hedge Funds
Three ex-DeepMind researchers, who previously developed an AI capable of defeating human poker champions, have pivoted that technology to financial markets — and the strategy is yielding significant returns. Their Prague-based AI venture, EquiLibre T
Sony, Warner Sue Anthropic Over Alleged Massive AI Copyright Infringement
Sony and Warner Music Group have initiated a landmark copyright lawsuit against AI firm Anthropic, alleging massive unauthorized use of copyrighted music for training the Claude model series. The complaint highlights Anthropic’s failure to secure lic
Qwen3.8-Flash Agent Officially Launched, Qwen Office Enters the Era of Abundant and Efficient Management
The Qwen Office Platform has officially released the Qwen3.8-Flash large model alongside a new standard mode, allowing all users to access this feature immediately. Leveraging enhanced model capabilities, users can now complete various office tasks w
Related Special Topic Recommendations
Comments (0)
0/500

On July 6th, the AI field saw a major milestone as Meituan officially announced the open-source release of its foundational large model, LongCat-2.0. Industry experts regard this move as a significant breakthrough in applying domestic computing power to large-scale AI models.
As the first foundational large model in the industry trained and inferred entirely on domestic computing chips, LongCat-2.0 demonstrates impressive technical strength. Officially released on June 30th, both training and inference were successfully completed on a domestic computing cluster consisting of 50,000 cards, fully showcasing the stability and high performance of domestic chips when handling trillion-parameter-level models.
To promote technological inclusivity and ecosystem growth, Meituan adopted a comprehensive open-source strategy. On July 6th, the model weights, inference engine, and detailed technical documentation for LongCat-2.0 were simultaneously released on major developer platforms such as GitHub, Hugging Face, and ModelScope. Notably, the model is distributed under the MIT open-source license, allowing developers and companies to use it for commercial purposes free of charge, significantly lowering the barrier for small and medium-sized enterprises to adopt cutting-edge AI technology.
The domestic chip ecosystem responded quickly. On the same day Meituan announced the open-source release, three domestic chip manufacturers—Huawei Ascend, MoLeLineage, and Muxi Technologies—jointly announced that they had completed inference adaptation for the LongCat-2.0 model. This collaborative effort not only validates the model's compatibility with domestic hardware but also marks the rapid maturation of the integrated hardware-software ecosystem combining domestic chips with domestic large models.
This open-source initiative is not only a concentrated demonstration of technical capability but also provides a solid technical foundation for China's AI industry's self-reliance and control through the construction of a complete open-source ecosystem. As more developers contribute to optimizing and applying LongCat-2.0, the voice of domestic AI computing power in international competition is expected to grow further.
DeepMind Trio Behind Poker AI Now Generating Returns for Quant Hedge Funds
Three ex-DeepMind researchers, who previously developed an AI capable of defeating human poker champions, have pivoted that technology to financial markets — and the strategy is yielding significant returns. Their Prague-based AI venture, EquiLibre T
Sony, Warner Sue Anthropic Over Alleged Massive AI Copyright Infringement
Sony and Warner Music Group have initiated a landmark copyright lawsuit against AI firm Anthropic, alleging massive unauthorized use of copyrighted music for training the Claude model series. The complaint highlights Anthropic’s failure to secure lic
Qwen3.8-Flash Agent Officially Launched, Qwen Office Enters the Era of Abundant and Efficient Management
The Qwen Office Platform has officially released the Qwen3.8-Flash large model alongside a new standard mode, allowing all users to access this feature immediately. Leveraging enhanced model capabilities, users can now complete various office tasks w





Home






