Janus

model paper dataset
DeepSeek
Unified autoregressive framework handling both multimodal understanding and visual generation (DALL-E style) in one model. Includes Janus and Janus-Pro versions with training data.
GitHubHuggingFace

Outputs 4

Janus & Janus-Pro

model
HuggingFace

Variants

Name Parameters Notes
Janus — Released Oct 2024
Janus-Pro — Updated through Jan 2026

Janus: Decoupling Visual Encoding for Unified Multimodal Understanding and Generation

paper

Unified autoregressive framework for multimodal understanding and visual generation.

Paper (arXiv)

arXiv: 2410.13848

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

paper

Scaling multimodal understanding and generation with data and model improvements.

Paper (arXiv)

arXiv: 2501.17811

Janus-Pro Training Data

dataset

72M descriptive mix of real and synthetic multimodal data for Janus-Pro training.

GitHubPaper (arXiv)
multimodalgenerationopen-weight