Replicate

Replicate is a cloud platform that lets developers run thousands of open-source AI models, including image, video, audio and language models, through a single hosted API. You push a prompt or file, Replicate handles the GPUs, scaling and cold starts, and bills only for the seconds your model runs. Teams also fine-tune models on their own data and deploy custom versions. It is popular with startups and indie builders who want production AI features without managing machine learning infrastructure.