AnalysisVisual AIOctober 18, 2026

Reddit user shares MOSS image model gallery at cfg 4, 39 steps

Read original source →reddit.com

Images generated with the MOSS image base model at cfg 4 and 39 steps, euler/sgm uniform sampler, at 1.33–1.67MP in 1:1, 3:4 and 4:3 ratios. The poster advises skipping the SFT model, calling it over-RL-trained with too much AI aesthetic and noise.

1 source

More stories today

Open the live feed