Skip to main content
A deployment serves a checkpoint or a base model behind an OpenAI-compatible endpoint. It runs in Monte Cloud, or in the team’s own account with BYOC.
Each deployment has replicas that scale with traffic, on a schedule, or to zero, and it keeps versions for traffic splits and rollback. Many LoRA adapters can share one base deployment. Callers send an API key, and the endpoint takes chat completions with streaming and tool calls. A deployment is billed until it is deleted or scaled to zero. A deployment also collects each request and response as a trace for continual learning. --traces off keeps no content.

In This Section