Multimodality · 2025

Janus-Pro: Unified Multimodal Understanding and Generation with Data and Model Scaling

Xiaokang Chen, DeepSeek-AI

An open autoregressive framework that decouples visual encoding for understanding versus generation, letting one model both interpret and produce images competitively, a template for unified multimodal LLMs.

Editorial record

Plain-language summary

An open autoregressive framework that decouples visual encoding for understanding versus generation, letting one model both interpret and produce images competitively, a template for unified multimodal LLMs.

Knowledge graph

Relationships

Descendants

  • Depends onEvidence: Strongly supported

    Attention Is All You Need

    Builds on the multimodal lineage in the archive

    freshness sweep 2026

Source record

Provenance

Record ID
P-622
Record created
2026-07-13
Last reviewed
2026-07-14
Record version
2

Citation caveat: Citation metadata is approximate and marked unverified in the source dataset.