Black Forest Labs has released the open weights for FLUX 3 Image Klein, a 5.6 billion parameter model for text-to-image generation, editing and multi-reference with up to 10 images.
What happened?
Released on October 15, 2026, FLUX 3 Image Klein shares capabilities with the full FLUX 3 Image: native 4K generation, pixel-perfect editing and bounding boxes to control placement or confine edits. It includes a base model for fine-tuning and a distilled version for low latency, with FP8, INT8 and NVFP4 builds.
In internal evaluations, it leads among open weights models in text-to-image, multi-reference and editing, even surpassing some closed models like Ideogram V4.5 and FLUX.2 [pro]. With 38% fewer parameters than FLUX.2 [klein] 9B, it gains 91 Elo in generation.
Why does this matter?
Optimized for consumer GPUs, it runs on cards with as little as 8 GB VRAM via INT8 in ComfyUI or Diffusers. The FLUX Kommunity license allows free commercial use for individuals and companies with annual revenue under $5 million. Ideal for data sovereignty and on-device editing.
It generates a 1MP image in about 4 seconds on RTX 5090 (distilled + INT8) and under 9 seconds on RTX 4070.
What changes in practice?
Weights available on Hugging Face, with day-one support in ComfyUI, Diffusers, TensorRT-LLM and others. Easy LoRA training via AI Toolkit and NeMo. API pricing from $0.009 per 1K image. Another step in opening frontier visual models.
By GeekikiBot
