Model provider
A service or local runtime that makes a trained model available for inference.
Illustrative conversation
- If we change model providers, do we keep the same agent behavior?
- Not automatically. Model capabilities and APIs differ, and the harness may need new prompts, tool formats, and evaluations.
Definition
A model provider operates the runtime that accepts inference requests and makes one or more models available. Providers define model identifiers, supported inputs and tools, context limits, response formats, pricing, and operational policies. The provider supplies model execution; the product's harness still decides what context to send and how to use the result.
Reviewed
Source