Meta: Muse Image

Muse Image is an image generation model from Meta, listed on api14 under the model id meta/muse-image. Meta describes it as an agentic image generation model, which means it does not simply map a prompt straight to pixels. Instead, it reasons before it renders, breaking a request down before producing output. The model sits in the image category and is identified by its provider release date of 2026-08-26.

On the input side, Muse Image accepts text and images. Output is image only. The context window is 65,536 tokens, and the maximum output is 58,982 tokens. Supported request parameters are max_tokens, repetition_penalty, temperature, top_k and top_p, so you have the usual sampling controls to steer how deterministic or varied the generated images are.

Because it takes both text and reference images as input, the practical uses follow two paths. Text-to-image work covers concept art, illustrations, marketing visuals and other assets generated from a written description. Image editing and transformation work covers changes to an existing image supplied as a reference, which is useful when you need variations on a design, adjustments to a photo, or edits that must stay close to an original. The agentic behaviour, where the model reasons before rendering, is most relevant when a request is complex or multi-part and a single pass would likely miss part of the intent.

Calling it through api14 works the same way as any other model on the gateway. Send a request to the OpenAI-compatible endpoint with the model field set to meta/muse-image. The request parameters you can set are max_tokens, repetition_penalty, temperature, top_k and top_p. Include your text prompt, and add reference images in the input when you want editing rather than generation from scratch. Any OpenAI-compatible client library will work without special handling beyond pointing it at the api14 base URL and using this model id.

One practical consideration is that the output modality is image only, so this model will not return text answers, code or structured data. If your task is text generation, reasoning over documents or anything that expects a text response, a text model on api14 is the better fit. Muse Image is the right choice when the deliverable is an image, whether generated from a prompt or edited from a reference, and you want a model that works through the request before it renders.