WHAT IT IS
Replicate is a platform introduced in 2019 that lets users run community-uploaded models through an API. Models are packaged as Cog containers, and the platform handles building, running and scaling, charging by usage. It addresses the difficulty of quickly deploying arbitrary models as callable services.
Why it matters
With container packaging and usage-based billing, it reduced "run someone else’s trained model" to a single API call.
Key specs
- Packaging
- Cog containers
- Billing
- Charged by usage