LongCat-Next: Lexicalizing Modalities as Discrete Tokens

TL;DR

LongCat-Next introduces a unified discrete autoregressive framework for multimodal modeling, enabling seamless integration of text, vision, and audio with hierarchical tokenization and strong benchmark performance.

Contribution

It presents Discrete Native Autoregressive (DiNA) and dNaViT, pioneering a shared discrete space for multimodal data, advancing beyond language-centric systems.

Findings

Achieves strong performance across multimodal benchmarks.

Addresses the long-standing performance ceiling in vision understanding.

Provides open-source tokenizers and models for community research.

Abstract

The prevailing Next-Token Prediction (NTP) paradigm has driven the success of large language models through discrete autoregressive modeling. However, contemporary multimodal systems remain language-centric, often treating non-linguistic modalities as external attachments, leading to fragmented architectures and suboptimal integration. To transcend this limitation, we introduce Discrete Native Autoregressive (DiNA), a unified framework that represents multimodal information within a shared discrete space, enabling a consistent and principled autoregressive modeling across modalities. A key innovation is the Discrete Native Any-resolution Visual Transformer (dNaViT), which performs tokenization and de-tokenization at arbitrary resolutions, transforming continuous visual signals into hierarchical discrete tokens. Building on this foundation, we develop LongCat-Next, a native multimodal model that processes text, vision, and audio under a single autoregressive objective with minimal modality-specific design. As an industrial-strength foundation model, it excels at seeing, painting, and talking within a single framework, achieving strong performance across a wide range of multimodal benchmarks. In particular, LongCat-Next addresses the long-standing performance ceiling of discrete vision modeling on understanding tasks and provides a unified approach to effectively reconcile the conflict between understanding and generation. As an attempt toward native multimodality, we open-source the LongCat-Next and its tokenizers, hoping to foster further research and development in the community. GitHub: https://github.com/meituan-longcat/LongCat-Next