Active

Replicate

Hosted APIs for running, fine-tuning, and deploying machine-learning models.

  • Catalog of public and proprietary models
  • API predictions and model-specific inputs
  • Dedicated deployments for custom scaling needs

Replicate runs models through a hosted API. Its catalog includes community models and proprietary services, with each model page describing inputs, versions, and pricing. Dedicated deployments provide additional control for selected workloads.

How pricing works

Some models charge for hardware running time; others charge by input or output, such as tokens or generated images. Dedicated deployments have their own billing behavior. Use the selected model's pricing and license rather than a platform-wide per-request estimate.

Sources

From the archive