Meituan open-sources trillion-parameter LongCat-2.0

Meituan has open-sourced LongCat-2.0, a trillion-parameter AI model trained and deployed entirely on Chinese computing infrastructure.

Meituan releases LongCat-2.0 to the public. The food delivery and services giant made the trillion-parameter model available as open source on July 8, 2026. LongCat-2.0 was developed completely within China's domestic AI stack.

The release transforms the previously anonymous Owl Alpha project into a prominent benchmark. It now serves as a visible test case for the viability of locally built AI models and supporting hardware.

Model relies exclusively on domestic infrastructure. Training and deployment of LongCat-2.0 used only Chinese computing resources and chips. No foreign hardware or software components were involved in the core pipeline.

This approach aligns with broader national efforts to reduce dependence on overseas technology. The model's scale, at one trillion parameters, places it among the largest publicly disclosed Chinese AI systems.

Release signals growing momentum in China's AI ecosystem. Meituan's move comes as domestic AI development accelerates amid supply chain and geopolitical constraints. LongCat-2.0 demonstrates that large-scale models can be built and run without reliance on restricted foreign platforms.

The open-sourcing invites wider participation from researchers and developers inside China. It positions the model as both a technical achievement and a strategic statement on self-reliant AI capabilities.