1 min read
Meituan Releases LongCat-2.5 AI Model Trained on Chinese Chips
The model has 1.6 trillion total parameters, activates about 48 billion per token and processes visual input alongside text.

Chinese delivery and local services company Meituan has released LongCat-2.5, a 1.6 trillion-parameter AI model trained entirely on domestic computing hardware, adding visual processing to its Chinese-chip AI effort.
LongCat-2.5 activates about 48 billion parameters for each token it processes, a fraction of its total capacity. It converts visual input into discrete tokens that the model can handle alongside text.
Its agent features are designed for extended tasks involving complex codebases and complete software projects.
Meituan has also launched 妙手, an AI workspace for merchant analysis and store management.