Black Forest Labs: FLUX 3 Image Klein

FLUX 3 Image Klein is an image generation and editing model from Black Forest Labs. It is an open-weights model, which means the weights are published rather than kept behind a hosted-only service. Black Forest Labs is the provider behind the FLUX family, and this entry sits in the image category on api14.

The model takes text and image inputs and produces image output. That combination covers both text-to-image generation and editing driven by reference images. The context window is 46,864 tokens, and the maximum output is 42,177 tokens. Image requests expose several controls: resolution can be set to 768, 1K, 1.5K, 2K or 4K; aspect ratio accepts a wide set of values including 21:9, 16:9, 3:2, 4:3, 1:1, 4:5, 3:4, 9:16, 9:21 and auto; the number of images per request is fixed at 1 (minimum 1, maximum 1); and input references can range from 0 to 10.

Those capabilities point to a few practical patterns. With zero reference images, the model works as a straightforward text-to-image generator, and the aspect ratio and resolution enums let you target a specific frame shape and output size, from 768 up to 4K. With one or more references, up to ten, it becomes an editing and composition tool: you can supply reference images and ask for a result that keeps characters or subjects consistent across the output. The provider description specifically calls out multi-reference editing that preserves characters and subjects, so projects with recurring characters, product shots or style references are the natural fit. The single-image output per request suits workflows where each call produces one asset to review or pass downstream.

You call it through api14's OpenAI-compatible API using the model id "black-forest-labs/flux-3-image-klein". Because the interface follows the OpenAI format, existing OpenAI client libraries and request shapes work with a change of base URL and model id. The supported parameter for this model is seed, which lets you reproduce or vary a generation deterministically. The image-specific controls listed above (resolution, aspect ratio, n and input references) are part of the request as well. Since n is capped at 1, batch generation means issuing multiple requests rather than asking for several images at once.

The main constraints to weigh are the fixed single-image output and the modest parameter surface, which is limited to seed plus the image controls. If your task is text-only, or if you need a model that returns text rather than images, this is not the right entry. If you need several images from one call, you would need to send multiple requests. For text-to-image and reference-based editing within the listed resolutions and aspect ratios, it is a direct fit.