Replicate
Run thousands of community-published models, image, audio, video and language, behind one API without hosting any of them yourself.
Tested on GPT-5 ·
SERVICE · Usage-based, pay per second of compute
The catalogue is the point: models other people packaged and published, all callable through one API whatever they run on underneath. To try a model someone else fine-tuned, or to use image, audio and video models outside the usual text APIs, it is usually the quickest path from found this on GitHub to a working API call.
Cold starts on less popular models can take seconds to a couple of minutes, because the hardware spins up on demand. Do not use it for latency-sensitive production traffic on an obscure model. It is fine for popular ones and for batch work.
Notes from the author
Look at a model recent run count before you depend on it. If nobody has run it in weeks, your first request hits a cold start, and you have no recent proof that it still works as documented.
For information only, not professional advice. You use it at your own risk and we are not liable for any loss or damage. See the disclaimer.
Tagged