The Temperature property returns or changes the sampling
temperature used when generating responses from the current model.
Temperature controls the amount of randomness used when selecting
tokens during text generation. Lower values generally produce more
predictable and consistent responses, while higher values can produce
more varied and creative output.
A temperature value of 0.0 requests the most deterministic
response possible from the model. This is typically recommended for
tasks that require stable or structured output, such as code
generation, classification, data extraction, or responses that must
follow a specific format. Higher values, such as 1.0 or above, may
produce more diverse wording, alternative phrasing, or creative
responses, but can also increase the likelihood of inconsistent or
inaccurate output.
Most providers support temperature values in the range of 0.0 to
2.0, although the exact supported range and behavior may vary between
providers and models. Some models may ignore this property entirely or
internally restrict the range of accepted values. If an unsupported
value is specified, the provider may clamp the value to a supported
range or reject the request.
Changing the temperature only affects subsequent requests and does
not modify existing messages or responses stored in the conversation
history.
In most cases, applications should adjust only the temperature
setting and leave advanced sampling parameters such as TopP
and TopK at their default values unless the application has
specific requirements and understands how those settings affect model
behavior.
The default temperature value depends on the selected provider and
model. Some providers may use different default values for reasoning
models, coding models, or creative writing models.