Replicate is joining Cloudflare (replicate.com)

Replicate announced it is joining Cloudflare and will continue to operate as a distinct brand while integrating with Cloudflare’s Developer Platform. The Replicate API and existing models will remain unchanged, so current apps keep running. Replicate brings its open-source Cog format (a standard for packaging models) and a hosted platform for publishing and running models via API — essentially low-level primitives that make it easy for developers to treat models like deployable services. The move matters because it pairs Replicate’s model runtime and developer-facing abstractions with Cloudflare’s global network and platform services. Expect faster, more scalable model hosting and tighter integration with Workers (for agents and glue code), Durable Objects (state management), R2 (file storage), and WebRTC (media streaming). That combination enables better orchestration, edge or real-time inference, and production-grade GPU/cluster scaling with minimal migration friction. For the AI/ML community, this is a step toward a "distributed operating system for AI" — easier deployments, lower operational overhead, and a path to become a default platform for building AI apps.