A image-to-image workflow to benchmark FLUX 3 KLEIN and FLUX 2 KLEIN 9B.
Performance (1024x1024): About 35 sec with a 5070ti, 70 sec with a 5060ti (2nd gen/warm).
FLUX 3 KLEIN: int8.convrot
FLUX 2 KLEIN 9B: FP8
LLM with vision: QWEN 3.5
This workflow generate a big image with the 2 images and the model's name on it. So you can easily share your results on social networks.
