Qwen released Qwen-Image-2.1-PE-I2I on Hugging Face
Qwen released Qwen-Image-2.1-PE-I2I on Hugging Face under the Qwen Research License Agreement. It is a fine-tuned Qwen3.5-VL 9B model that rewrites vague editing instructions into precise prompts for Qwen-Image-2.1. Qwen-Image-2.1 has a 7B visual generation component (32 Single-Stream DiT layers), supporting up to 10 reference images, native RGBA transparency, and local edits via circles, painted annotations, or masks.
China context
- Original name
- Qwen-Image-2.1-PE-I2I
- Outside China
- Open weights · huggingface.co
- Claims
- Company-reported; not yet independently evaluated
- For builders
- Developers outside China can download the model weights from Hugging Face and integrate the prompt rewriting model into image editing pipelines using Diffusers.
- For investors
- The release of auxiliary models alongside the main generation model indicates Qwen's strategy to build a full open-weights image editing stack, which may increase adoption and ecosystem lock-in.
Qwen released Qwen-Image-2.1-PE-I2I, an image editing prompt rewriting model for Qwen-Image-2.1, on Hugging Face. The model is a fine-tuned Qwen3.5-VL 9B that takes vague editing instructions and input images to produce precise prompts for downstream image editing. Qwen-Image-2.1 is a unified text-to-image generation and image editing model with 7B parameters in its visual generation component, supporting up to 10 reference images, native transparency, and versatile editing.
Qwen-Image-2.1-PE-I2I is a fine-tuned Qwen3.5-VL 9B model that outputs a JSON object with a rewritten prompt and aspect ratio information after a reasoning block. It supports multiple input images and integrates with Diffusers. Qwen-Image-2.1 uses a lightweight architecture with mixed-granularity attention and prefix KV cache reuse, and its visual generation component has 7B parameters across 32 Single-Stream DiT layers.
Qwen continues to expand its open-weights image generation ecosystem by releasing auxiliary models like prompt rewriters alongside the main generation model. This modular approach allows users to improve editing workflows without retraining the core model.
For developers, Qwen-Image-2.1-PE-I2I can streamline image editing workflows by automatically generating precise prompts from vague instructions, reducing manual prompt engineering. For investors, the release signals Qwen's commitment to building a comprehensive open-weights image generation ecosystem.
Observable next signals include adoption of Qwen-Image-2.1-PE-I2I in image editing pipelines, community feedback on prompt rewriting quality, and potential releases of additional auxiliary models for Qwen-Image-2.1.