Black Forest Labs has released the open weights for FLUX 3 Image Klein, a 5.6-billion-parameter image model that the German lab calls its most capable open-weights image model to date. It inherits the headline features of the full FLUX 3 Image model — native 4K output, pixel-perfect editing and bounding-box layout control — plus multi-reference generation from up to 10 reference images, all in a package small enough to run on consumer hardware.
Smallness is the point. The INT8 build runs on NVIDIA GeForce cards with as little as 8GB of VRAM through ComfyUI or Diffusers — meaning a standard gaming laptop can generate and edit images locally. BFL reports the distilled INT8 model generates a 1MP image in about 4 seconds on an RTX 5090 and under 9 seconds on an RTX 4070. The release ships two ways: a base model for fine-tuning and research, and a step-distilled model for low-latency generation, with FP8, INT8 and NVFP4 quantized builds developed with NVIDIA.
BFL's internal evaluations put Klein first among leading open-weights image models in text-to-image, multi-reference generation and editing — ahead of larger models like Qwen-Image 2.1 (7B) and Krea 2 (12.9B), and a reported 91 Elo gain over its own FLUX.2 klein with 38% fewer parameters. As always, treat company benchmarks as claims until independent testing lands. Early community reports also note weaker performance on generating people, apparently from aggressive scrubbing of training data.
For businesses, the license matters. The weights ship under the FLUX Kommunity License: free commercial use for companies under $5 million in annual revenue, with larger organizations needing a separate arrangement. The hosted BFL API starts at $0.009 per image at 1K resolution. Klein follows September's open-weights release of FLUX 3 Action for robotics — BFL says more FLUX 3 family releases are coming.