LongCat-2.5 Debuts With 1.6T Parameters and Native Visual Tokens

LongCat released LongCat-2.5, a 1.6 trillion parameter multimodal mixture-of-experts model with native visual token processing and enhanced agentic coding for complex long-horizon tasks. It integrates visual and textual inputs via discrete visual tokens and uses reinforcement learning on large codebases, linking to chat and API platforms.