flux-dev corresponds to Black Forest Labs' FLUX.1 [dev], an open-weight image model trained with guidance distillation that excels at turning text descriptions into visual works. On this platform, it supports text generation and editing existing images, making it suitable for concept design, illustration drafts, and visual concept experimentation, organizing the creative workflow through prompts, aspect ratios, and quantity.
Clarify capacity, inputs and outputs, and invocation methods before selecting a model.
The architecture and training method are native characteristics of FLUX.1 [dev], while sizing, editing, and task management are capabilities provided by this platform.
Learn what flux-dev can bring to your work.
FLUX.1 [dev] emphasizes prompt adherence, making it suitable for putting the subject, environment, composition, and lighting into a single description. When creating, you can first define the main subject, then add materials and visual style, comparing results iteration by iteration; how prompts are phrased affects the generated image, so avoid piling all requirements into mutually contradictory keywords.
In addition to starting from text, flux-dev can also accept existing image links and editing instructions through edit mode, bringing drafts into the next round of creation. It is suitable for trying new visual directions or adjusting image content, but editing does not mean precise local replacement; elements that need to be preserved should be clearly described, and actual results should be checked one by one.
Square, portrait, and landscape sizes allow the same theme to be developed for different display placements. Generation tasks can use count to request multiple candidate images, then retrieve works from the result links; asynchronous queries or completion callbacks make it easier to integrate the generation process into background tasks without keeping the frontend waiting on the same connection.
Start with specific tasks to find where the model can be effective.
Enter character traits, scene atmosphere, color direction, and composition requirements to generate concept images or illustration drafts for discussion. It is suitable for exploring multiple visual directions before formal drawing, then having designers select and refine them; the deliverables are image candidates for review, not automatically completed final design specifications.
Describe the background, materials, lighting, and placement around a product theme to generate scene visual proposals; you can also submit existing images to try editing. It is suitable for comparing presentation atmospheres and composition directions, while the shapes, logos, and details of real products still need verification, and generated images should not be directly treated as accurate product photographs.
Content applications can turn users' theme descriptions into prompts, choose landscape or portrait formats, and submit generation tasks, then display image links for users to choose from after completion. It is suitable for article illustrations, social content, and event visual drafts; when accurate titles or brand text are needed, typography can be left to later design stages.
Choose based on task complexity, input materials, and expected results.
If the main task is validating prompts, exploring styles, and creating concept drafts, flux-dev is a suitable starting point. FLUX.1 [dev] and FLUX.1 [pro] are different models, and the official positioning rates the output quality of the contemporary pro higher; for finished tasks with stricter detail requirements, try the same concept with each separately rather than treating dev as another name for pro.
When you need to try modifications to an existing image, you can use flux-dev's edit mode; if the task focuses on contextual editing around the original image, Flux Kontext should be evaluated first. Do not equate dev's editing entry point with Kontext's dedicated editing capabilities, and do not apply FLUX.2's new features to this generation of models.
From a small-scale task to formal integration.
Define the goal, required inputs, and output requirements, using real business examples as a starting point.
Open the trial page, confirm the parameters supported by this entry point, then submit a small-scale task to review the results.
Keep the full model ID, use the request format specified in the documentation, and confirm billing rules on the Pricing page.
Before formal use, understand the output quality and capability scope.
Answers to common questions about using flux-dev.
flux-dev is the invocation ID for selecting FLUX.1 [dev] on this platform, and the model is developed by Black Forest Labs. It belongs to the FLUX.1 series; it is not Flux Kontext or FLUX.2, nor does it automatically gain the capabilities of those models by sharing an image interface.
Submit model=flux-dev, action=generate, prompt, and size to POST /flux/images, and authenticate using a Bearer Token. Results are returned as JSON, and images are obtained through image_url in data; set count when multiple candidates are needed.
Yes. Use action=edit and submit image_url, an editing prompt, and size. Editing is suitable for further exploring changes to an existing image, but it does not guarantee that unspecified areas will remain completely unchanged; count is not used for editing tasks, and important details should be checked one by one after the results are returned.
For square images, use 1024x1024; for portrait images, use 1024x1792; for landscape images, use 1792x1024. You can also set the canvas through the image aspect ratio. It is recommended to determine the orientation based on the display location first, then describe the subject position and negative space in the prompt, rather than changing only the size while ignoring composition.
Yes. You can set async=true to obtain a task_id first and then query the task result; you can also provide callback_url to receive a result notification upon completion. Applications should distinguish between a submitted task and a completed image, and display the image only after obtaining its link, to avoid treating a task ID as a generated image result.
Model information · Updated: 2026-10-01. Please see the API and pricing sections for invocation parameters and billing rules.
Start with a clear goal, and judge from real results whether it suits your work.
This is only a basic call example. See the full docs for more parameters and advanced usage.
View Docs