LongCat-2.5 Released: A Native 1.6 Trillion-Parameter Multimodal Model
The LongCat team has released LongCat-2.5, a native multimodal model designed for long-horizon agent tasks. With 1.6 trillion total parameters and approximately 48 billion activations per token, it builds on LongCat-2.0 by enhancing visual understanding and agentic coding. A trial version is now available on its official website and accessible via API, while the GitHub and HuggingFace versions are not yet open.
LongCat-2.5 Released: A Native Multimodal Model with 1.6 Trillion Parameters
The LongCat team has released LongCat-2.5, a native multimodal model designed for long-context agent tasks. It boasts a total of 1.6 trillion parameters, with approximately 48 billion activations per token. Building upon LongCat-2.0, it enhances visual understanding and agentic coding capabilities.
The model encodes visual signals into multi-level discrete tokens, which are then processed alongside text, enabling it to perform tasks such as visual programming. It is now available for trial on its official website and can be accessed via an API. The GitHub and HuggingFace versions are not yet publicly available.
Source: LongCat Official Blog