The Qwen team has announced Qwen-Image-2.0, a model that promises to significantly improve image generation and editing.

Key Features

  • Reduced size: The model has a size of 7B parameters, a significant drop compared to the 20B of the previous version. This makes it more suitable for running on less powerful hardware.
  • Unified functionality: Qwen-Image-2.0 offers image generation and editing in a single pipeline, eliminating the need for separate models.
  • High resolution: It natively supports 2K (2048x2048) images with realistic texture rendering.
  • Advanced text rendering: Handles text rendering from prompts up to 1000 tokens, paving the way for the creation of infographics, posters and other visual materials.
  • Comic generation: Ability to generate multi-panel comics (4x6) with consistent characters.

Availability

Currently, Qwen-Image-2.0 is available via API on Alibaba Cloud (invite beta) and via a free demo on Qwen Chat. The release of the model weights is expected soon, following the strategy adopted with Qwen-Image v1.

The reduction in model size to 7B is a particularly interesting aspect for those who want to run models locally. The previous 20B version was already popular in environments like ComfyUI, and a lighter version with improved functionality could further expand its adoption.