Reasoning interface

gpt-5 and o-series models only Configuration options for reasoning models.

Properties

context	Controls whether reasoning items from the current turn or all turns are included in later model context; `auto` lets the model choose.
effort	The reasoning effort requested for the response. Lower effort can reduce latency and reasoning token usage; supported values depend on the model.
generate_summary	Legacy setting for the level of detail in the model's reasoning summary. Use `summary` for new configurations.
mode	Controls the reasoning execution mode for the request. When returned on a response, this is the effective execution mode.
summary	The level of detail requested for the model's reasoning summary, or `auto` to let the model choose.

Controls whether reasoning items from the current turn or all turns are included in later model context; auto lets the model choose.

context?: "auto" | "current_turn" | "all_turns"

"auto" | "current_turn" | "all_turns"

The reasoning effort requested for the response. Lower effort can reduce latency and reasoning token usage; supported values depend on the model.

effort?: ReasoningEffort

Legacy setting for the level of detail in the model's reasoning summary. Use summary for new configurations.

generate_summary?: "auto" | "concise" | "detailed"

"auto" | "concise" | "detailed"

Controls the reasoning execution mode for the request. When returned on a response, this is the effective execution mode.

mode?: string

string

The level of detail requested for the model's reasoning summary, or auto to let the model choose.

summary?: "auto" | "concise" | "detailed"

"auto" | "concise" | "detailed"