Beta

AI Foundry is in beta. We are actively shaping the product, so things may change as we iterate. Your feedback is welcome.

Model

A Model is a catalog resource that wraps the configuration needed to connect to a large language model (LLM) provider. It decouples the LLM choice from the agents that use it: agents reference a model by name, so swapping the underlying model (or updating API credentials) requires changing only the Model resource without touching any Agent.

.

Model reference

Field	Required	Description
Title	Yes	Display name shown in the UI and in the agent creation form's model picker.
Name	Yes	Unique identifier. Must match the value used in Agent.spec.model.
Description	Yes	Short description of the model's capabilities or intended use.
Type	Yes	Provider identifier string. Common values: openai, anthropic, azure, ollama/<model>, mistral. Passed to the LLM client unchanged.
Provider	Yes	Provider name displayed in the UI model picker (e.g. openai, anthropic, ollama).
Secret	No	Name of the environment variable that holds the API key. The key is never stored in the catalog; only its variable name is.
URL	No	Base URL of the LLM API endpoint. Defaults to the provider's public endpoint when omitted. Must be a valid URL.
Context Window	No	Maximum context window size in tokens. Informational; surfaced in the model detail view.
Supports Tools	No	Set to true if the model supports tool/function calling. Used by the AI Playground to determine which agent features are available. Defaults to false.
Supports Vision	No	Set to true if the model accepts image inputs. Defaults to false.
LiteLLM	No	When true, requests are routed through a LiteLLM proxy, enabling a unified interface for any provider LiteLLM supports. Defaults to false.
Arguments	No	A free-form JSON object with default parameters applied to every request (e.g. temperature, max_tokens, top_p). Individual agents can override these via Agent.spec.model_arguments.

See also

• Agent: resources that reference a model.