New Version of the Flux2-Klein-9B-True-V3 Model

🎨 New Version of the Flux2-Klein-9B-True-V3 Model

A fine-tuned version of Black Forest Labs' Flux.2-klein-9B has been released, optimized to improve aesthetics, composition, and instruction following. By utilizing INT8 quantization and integration with the ComfyUI-INT8-Fast extension, generation speed increases by 1.5–2x compared to the FP8 format.

🌍 The emergence of efficient quantization methods (INT8 with ConvRot) allows heavy models (9B parameters) to run on consumer hardware with minimal loss in quality.

👤 You can use the high-quality Flux2-Klein model at much higher speeds on NVIDIA RTX 30 series graphics cards and above using ComfyUI.

Source 1: https://huggingface.co/wikeeyang/Flux2-Klein-9B-True-V3 Source 2: https://github.com/BobJohnson24/ComfyUI-INT8-Fast