Qwen: Qwen Image 3

Qwen Image 3 is an image generation and editing model from Qwen, listed on api14 under the model id qwen/qwen-image-3. It sits in the image category and is described by its provider as a unified model, meaning generation and editing are handled by the same model rather than separate endpoints. If you are already working with other Qwen models, this one extends that family into visual output.

The model accepts text and image inputs and produces image output. Its context window is 65536 tokens, with a maximum output of 58982 tokens. Image generation is controlled through dedicated parameters: resolution can be set to 1K or 2K, and aspect ratio supports 1:1, 1:2, 1:4, 2:1, 2:3, 3:2, 3:4, 4:1, 4:3, 4:5, 5:4, 9:16 and 16:9. You can request between 1 and 6 images per call with n, and supply between 0 and 4 input references, which is what makes editing and reference-guided generation possible. Seed is also available as a boolean option. The provider notes precise rendering of text and details as small as 10px, which matters when your output needs legible labels, captions or fine graphic elements.

Those capabilities point to a few practical uses. Text rendering at small sizes suits poster and banner drafts, UI mockups, social cards and any asset where words must stay readable. The input reference range of up to 4 images makes it a fit for editing tasks such as restyling an existing image, combining elements from several references, or producing variations that stay close to a source. The aspect ratio list covers square, portrait, landscape and the very wide 4:1 and tall 1:4 formats, so you can target feed posts, stories, thumbnails and banners without post-processing crops. Generating up to 6 images in one request helps when you want options to choose from before committing to a direction.

You call the model through api14's OpenAI-compatible API using the model id qwen/qwen-image-3, so existing OpenAI client libraries work with a change of base URL and model name. The supported request parameters are max_tokens, presence_penalty, response_format, seed, temperature and top_p, alongside the image-specific controls for resolution, aspect ratio, n and input references. Note that seed appears both as a supported request parameter and as a boolean image parameter, so check how your client maps it. Because output is an image, response_format is worth setting explicitly in your request.

There is no information here about pricing, latency or rate limits, so check the live price table and your own testing before committing to production volumes. If your task is text generation or analysis rather than image output, this model is the wrong fit since it only returns images. For straightforward generation with no reference images, set input_references to 0 and rely on the text prompt alone.