Replicate

Replicate provides APIs for developers to run AI models, fine-tune them with their own data, and deploy custom models. Its catalog covers image and video generation, speech, music, image restoration, and language models. Developers can build on community-published models or bring their own code, while Replicate manages the infrastructure and scales capacity with demand. Its managed service charges for compute usage, giving teams a way to add AI features without operating GPU clusters themselves.

Founded in 2019 by Ben Firshman and Andreas Jansson, Replicate began with tools for making research models accessible to software developers. Firshman is its founder and CEO. Their foundational tool, Cog, packages models and dependencies into Docker containers, configures compatible GPU software, and generates validated inference APIs from Python input and output definitions. Those containers can run on a developer’s own infrastructure or on Replicate, separating portable model packaging from managed hosting.

Replicate reported 2 million sign-ups and 30,000 paying customers in December 2023, with the sign-ups accumulated over the preceding year and a half. Its business users included Unsplash, which used models to label its image catalog. Cloudflare completed its acquisition of Replicate on December 1, 2025, for $57.4 million in total purchase cash consideration, placing the model-serving platform alongside Cloudflare’s application compute, storage, and networking services.