>>105808503
yeah, trained on 41 tho
we now have v42, v42-detail-enhanced, and (v41 based) -staging cfg, large and RL
large - trained at 1024
cfg - cfg1 test
RL - i forget what it stands for but "the RL model is both CFG1 and low step model by design"
it can go down to 10 steps apparently @cfg1