DeepSeek
Parameter reference for the DeepSeek model class, including thinking mode controls.
The DeepSeek model provides access to DeepSeek's language models.
Parameters
| Parameter | Type | Default | Description |
|---|---|---|---|
id | str | "deepseek-v4-flash" | The ID of the DeepSeek model to use |
name | str | "DeepSeek" | The name of the model |
provider | str | "DeepSeek" | The provider of the model |
api_key | Optional[str] | None | The API key for DeepSeek (defaults to DEEPSEEK_API_KEY env var) |
base_url | str | "https://api.deepseek.com" | The base URL for the DeepSeek API |
reasoning_effort | Optional[str] | None | Controls thinking-mode effort: "low", "high", or "max". The provider defaults to "high"; request "max" explicitly. |
use_thinking | Optional[bool] | None | Controls thinking mode. None uses the model default (on for thinking-capable models). Set False for faster responses without thinking |
supports_native_structured_outputs | bool | False | Whether the model supports native structured outputs |
DeepSeek extends OpenAILike and accepts all of its parameters.
In thinking mode, the provider ignores sampling controls such as temperature, top_p, presence_penalty, and frequency_penalty. Inherited parameter acceptance does not change provider support.