LongCat-2.5: Native Multimodality for Agentic Tasks

45 points by pretext

LongCat-2.5 is a large model with 1.6 trillion parameters, fully trained using domestically produced chips. This development highlights advancements in native multimodality for agentic tasks, showcasing the capabilities of local technology in handling complex AI training processes.

Read the article

Discussion on Hacker News