Replicate

Replicate is an API for running open-source ML models in the cloud with no infrastructure to manage. You call the API with inputs and get back outputs — image generation (Flux, Stable Diffusion), video, audio, language models, and more, all on-demand. Developers use it to build apps that need model inference without spinning up GPUs, and for one-off runs of models too large or complex to self-host. The web UI lets you try any model instantly before writing a line of code.

Billing is usage-based per second of compute — popular image models cost fractions of a cent per run. No subscription required. Most often compared to fal.ai and Modal for serverless ML inference — Replicate's edge is the enormous catalogue of community-hosted models and the approachable interface for non-infra developers; fal.ai is faster for media generation; Modal gives more control for custom deployments.