Black Forest Labs: flux-3-image-klein

FLUX 3 Image Klein is an open-weight image model from Black Forest Labs, released on October 15, 2026. It accepts text and images and returns images, covering text-to-image generation, editing from a single reference image, and composition from up to 10 reference images. Bounding boxes on a 0 to 1000 grid steer the placement and scale of elements without hard-masking pixels, and the model can render text inside generated images. A prompt upsampler expands short requests into detailed scene captions and is on by default in the CLI, demo and ComfyUI; using the model without it significantly weakens results. The image transformer has 5.6B parameters and pairs with Qwen3-VL-4B-Instruct as a separate text encoder. A distilled checkpoint runs in 16 steps without guidance and backs the BFL API, while the undistilled Klein Base checkpoint uses 50 steps at CFG 3.5 for fine-tuning, LoRA training and research. Weights are 11.2GB in BF16 and 5.6GB in FP8, and an INT8 build runs with 8GB of VRAM on NVIDIA GeForce RTX GPUs via ComfyUI or Diffusers, generating a 1-megapixel image in about 4 seconds on an RTX 5090 and under 9 seconds on an RTX 4070. It suits creators and developers who need fast local or on-device image editing, multi-reference composition, and deployment on their own infrastructure for data sovereignty. Internal evaluations place it first for text-to-image, multi-reference generation and image editing, ahead of larger open-weight models and above Ideogram V4.5 and FLUX.2 pro. Non-commercial use is permitted under the FLUX Kommunity License v1.0, with commercial output exceptions for eligible users subject to conditions and restrictions; other uses may require a separate license from Black Forest Labs.