ModelSettings
type ModelSettings = object;Settings to use when calling an LLM.
This class holds optional model configuration parameters (e.g. temperature, topP, penalties, truncation, etc.).
Not all models/providers support all of these parameters, so please check the API documentation for the specific model and provider you are using.
Properties
Section titled “Properties”contextManagement?
Section titled “contextManagement?”contextManagement?: ModelSettingsContextManagement;Context-management strategies to apply when calling the model. This setting is available on OpenAI Responses requests, including server-side compaction. See https://developers.openai.com/api/docs/guides/compaction.
frequencyPenalty?
Section titled “frequencyPenalty?”frequencyPenalty?: number;The frequency penalty to use when calling the model.
maxTokens?
Section titled “maxTokens?”maxTokens?: number;The maximum number of output tokens to generate.
parallelToolCalls?
Section titled “parallelToolCalls?”parallelToolCalls?: boolean;Whether to use parallel tool calls when calling the model. Defaults to false if not provided.
presencePenalty?
Section titled “presencePenalty?”presencePenalty?: number;The presence penalty to use when calling the model.
preserveRawUsage?
Section titled “preserveRawUsage?”preserveRawUsage?: boolean;Whether to preserve a JSON-compatible snapshot of provider usage before the Agents SDK normalizes it. Defaults to false if not provided.
promptCacheOptions?
Section titled “promptCacheOptions?”promptCacheOptions?: ModelSettingsPromptCacheOptions;Controls implicit and explicit prompt caching for supported model providers.
promptCacheRetention?
Section titled “promptCacheRetention?”promptCacheRetention?: "in-memory" | "24h" | null;Enables prompt caching and controls how long cached content should be retained by the model provider. See https://platform.openai.com/docs/guides/prompt-caching#prompt-cache-retention for the available options.
providerData?
Section titled “providerData?”providerData?: Record<string, any>;Additional provider specific settings to be passed directly to the model request.
reasoning?
Section titled “reasoning?”reasoning?: ModelSettingsReasoning;The reasoning settings to use when calling the model.
retry?
Section titled “retry?”retry?: ModelRetrySettings;Runtime-only retry configuration for the model request.
store?
Section titled “store?”store?: boolean;Whether to store the generated model response for later retrieval. Defaults to true if not provided.
temperature?
Section titled “temperature?”temperature?: number;The temperature to use when calling the model.
text?: ModelSettingsText;The text settings to use when calling the model.
timeoutMs?
Section titled “timeoutMs?”timeoutMs?: number;Cooperative timeout in milliseconds for each model call. The timeout aborts only the current model call; a run-level abort signal still cancels the entire run. Must be greater than 0 and less than or equal to 2147483647 when provided.
toolChoice?
Section titled “toolChoice?”toolChoice?: ModelSettingsToolChoice;The tool choice to use when calling the model.
topP?: number;The topP to use when calling the model.
truncation?
Section titled “truncation?”truncation?: "auto" | "disabled";The truncation strategy to use when calling the model.