Flux 2 Klein 9B INT8 ConvRot Quantizations, for use in ComfyUI
FLUX.2 [klein] 9B-KV is an optimized variant of FLUX.2 [klein] 9B with KV-cache support for accelerated multi-reference editing. This variant caches key-value pairs from reference images during the first denoising step, eliminating redundant computation in subsequent steps for significantly faster multi-image editing workflows. Use the Flux KV Cache Node before inputting into your cfg guider or sampler of choice
Description
FLUX.2 [klein] 9B-KV is an optimized variant of FLUX.2 [klein] 9B with KV-cache support for accelerated multi-reference editing. This variant caches key-value pairs from reference images during the first denoising step, eliminating redundant computation in subsequent steps for significantly faster multi-image editing workflows.
FAQ
Comments (2)
Thanks for this.
Sometimes out of the blue it will ignore the img2img pose and eveyrhting and do portraits of what was in the orignal image, doing a re-run will fix it but its strange as i don't have that issue with default klein. any ideas?
Edit, seems to happen if i'm going too high in megapixels.
Thanks a lot for creating these. Would it be possible to do it for the qwen3 text encoder as well?