Black Forest Labs Releases FLUX.3 Image Klein
Black Forest Labs has released FLUX.3 Image Klein, a 5.6-billion-parameter open-weights model capable of text-to-image generation and advanced multi-reference image editing. The model utilizes Qwen3-VL-4B-Instruct as its text encoder and accepts a prompt along with up to 10 reference images to generate a single output image. This release marks a significant advancement in open-weights multi-modal AI, outperforming several larger open models and competing closed systems in internal evaluations. It provides developers and smaller businesses with powerful, accessible tools for complex image manipulation and generation tasks. Text-to-image generation, single-reference edits, and multi-reference compositions are all handled seamlessly through this unified 5.6B parameter model. Despite its relatively compact size, it manages complex image composition tasks efficiently by processing multiple inputs concurrently.
## BACKGROUND
Open-weights models provide researchers and developers with access to model parameters, enabling customization and self-hosting while lowering barriers to AI innovation. Text encoders like Qwen3-VL-4B-Instruct translate human language inputs into mathematical representations that multimodal AI models can understand and process.