Replicate

Replicate is a cloud API for running machine learning models and related workflows. The docs show you can run models from Node.js, Python, Google Colab, and other integrations, and the explore page focuses on models you can run in the cloud.

The pricing page says you only pay for what you use. Public models are billed either by the time they take to run or by input and output, while private models generally run on dedicated hardware and are billed for the time instances are online, with a special case for fast-booting fine-tunes.

The docs list several ways to work with the platform, including HTTP API access, client libraries, an OpenAPI schema, MCP server support, webhooks, and agent skills. They also document model creation, predictions, deployments, secrets, and rate limits.

The explore page highlights public and official models for image generation, video generation, text-to-speech, coding, and other model categories. The docs and pricing pages also show that users can deploy custom models with Cog.

The source confirms documentation for model running, fine-tuning, deployments, webhooks, billing, and security topics, but it does not provide a complete list of every supported framework, cloud provider, or runtime from the collected pages.