Skip to main content
Rase AI Intelligence Newsletter

Creative AI Model or AI service

Janus-Pro handles image questions and image generation in one model family

DeepSeek’s Janus-Pro can answer questions about an image or generate an image from a written description. The project provides downloadable weights in two sizes. [1][2]

Model access
Janus-Pro 1B and 7B weights are downloadable. [1][2]
License
MIT License for the repository code. DeepSeek Model License for use of the model weights. [1]

What the model does

DeepSeek released Janus-Pro on January 27, 2025 as an update to Janus. The project lists 1B and 7B checkpoints and describes separate visual representations for understanding and generation. [1][2]

Why it matters

A project can evaluate visual question answering and image generation within the same family. Each function has a different success measure, so comparisons should score them separately. [1][2]

Who can use it

Developers working on image search or design prototypes can evaluate either checkpoint. The official model card gives separate examples of understanding and generation. [2]

Key dates

  1. Janus-Pro introduced [1]

Ways to use it

Ask the model to describe an original visual and compare the answer with what the image actually shows. For generation, check the requested objects and layout against the result. [1][2]

Model limitations

The project uses separate visual representations for the two functions, and accuracy on one does not establish accuracy on the other. The repository code and model weights have different terms. [1][2]

Sources and verification

Catalog evidence checked How we verify records

  1. Primary source DeepSeek’s Janus-Pro release history, weights and licensing
  2. Primary source deepseek-ai/Janus-Pro-7B
  3. Other source Janus-Pro: Unified Multimodal Understanding and Generation with Data and Model Scaling

Related profiles

Creative AI

Seedance 2.5 (4K API and Flash)

ByteDance (Volcano Engine / BytePlus)

These tools generate video through developer services. The cited sources describe different variants and routes of access.

Medium confidenceEvidence checked

Creative AI

LongCat-Video-Avatar 1.6

Meituan (LongCat team)

It generates video of an avatar from supported inputs in the documented project.

High confidenceEvidence checked

Creative AI

MiniMax H3-Image

MiniMax

It generates images from prompts through the model architecture described by MiniMax.

High confidenceEvidence checked