Qwen-Image-Edit

Qwen-Image-Edit is an image editing model developed by Qwen, based on the 20B-parameter Qwen-Image architecture (Qwen2.5-VL + VAE Encoder). It extends Qwen-Image\'s text rendering capabilities to support precise image editing tasks, including both semantic and appearance modifications while preserving original image consistency.

Key features include: 

• Semantic Editing: Preserves overall semantic consistency during modifications (e.g., rotating objects, altering styles, or creating novel views). Examples include rotating a Capybara mascot 180 degrees or converting portraits into Studio Ghibli-style art. 
• Appearance Editing: Performs localized edits (e.g., adding a signboard with reflections, removing small objects, or modifying specific elements like a single letter\'s color). 
• Bilingual Text Editing: Directly edits Chinese and English text in images while retaining original font, size, and style. 
• Chained Editing: Supports iterative corrections, such as step-wise error fixes in calligraphy artworks using bounding box annotations.  

---

The model is a component of the image generation pipeline, consisting of:

• Text encoder: ~8.3B parameters,
• Transformer: ~20.4B parameters,
• VAE: ~126M parameters.

Total: ~28.8B parameters

---

The 'VRAM requirements' line is calculated based on the size of the weights in the specified precision, with an additional +10% coefficient for inference:
  model_parameters_count * bytes_per_parameter * 1.1 / 1024^3 
where model_parameters_count is number of model parameters, bytes_per_parameter is number of bytes used to store one parameter. For 16 bit = 2, 8 bit = 1, 4 bit = 0.5.

Announce Date: 17.08.2025
Parameters: 20B
Context: 128K
VRAM requirements: 14.8 GB using 4 bits quantization, 29.6 GB using 8 bits quantization, 59.1 GB using 16 bits quantization
Developer: Qwen
License: Apache 2.0

Use our pre-built public endpoints for free to test inference and explore Qwen-Image-Edit capabilities. You can obtain an API access token on the token management page after registration and verification.

Rent your own physically dedicated instance with hourly or long-term monthly billing.

We recommend deploying private instances in the following scenarios:

• maximize endpoint performance,
• enable full context for long sequences,
• ensure top-tier security for data processing in an isolated, dedicated environment,
• use custom weights, such as fine-tuned models or LoRA adapters.

Contact our dedicated neural networks support team at nn@immers.cloud or send your request to the sales department at sale@immers.cloud.