LongCat-Next:基于离散原生自回归范式的多模态统一建模
LONGCAT-NEXT多模态模型离散表示DINA视觉理解语音合成
美团 LongCat 团队提出 LongCat-Next 模型,通过离散原生自回归(DiNA)范式实现文本、视觉与音频的统一建模,在理解、生成与语音任务中表现优异。
Generated from longcat.chatMar 27, 2026
Oops! Chart Generation Failed
The data format seems to be invalid or incomplete. You can try to fix it manually in the editor.
Scroll to Zoom • Drag to Pan