@paytondev i assume for APIs? if you’re building an app it’s probably easier to implement an image generation feature with a model that spits out an image 100% of the time, vs one that might spit out text and hallucinate when meaning to generate an image