Meituan Releases and Open-Sources Native Multimodal Large Model LongCat-Next

Recently, Meituan released and fully open-sourced the native multimodal large model LongCat-Next. This model breaks the traditional patchwork architecture of large models centered on language, unifying images, speech, and text into homologous discrete tokens.

Through the pure 'next token prediction' paradigm, LongCat-Next makes vision and speech the 'native language' of AI, marking an important progress of Meituan's LongCat team in the field of physical world AI.