Meituan releases native multimodal LongCat-Next

On March 27th, Meituan released and fully open-sourced the native multimodal large model LongCat-Next and its core component - the discrete native resolution visual segmentation tool. This model breaks the traditional piecemeal architecture of current large models centered on "language" by unifying images, speech, and text into homogenous discrete Tokens. Through the pure paradigm of "next token prediction", LongCat-Next enables vision and speech to become the "native language" of AI.