Focused generation
Generate stable, focused answers or code with less variation.
AI & ML · Const
Temperature
Temperature controls sampling randomness in a completion. Values from 0 to 2 trade determinism for variety: lower is more focused, higher is more random.
How it works
Temperature controls randomness during sampling. Lower values make output more focused and deterministic, while higher values increase variation; it accepts values from 0 to 2.
Syntax
temperature: number | nullParameters
| Name | Type | Description |
|---|---|---|
temperature | number or null | Sampling temperature, from 0 to 2. Higher values increase randomness; lower values make output more focused and deterministic. |
Returns
The generated completion, including its text and associated metadata.
Key points
- Set temperature to a number from 0 through 2.
- Higher settings produce more varied and less predictable output.
- Lower settings favor focused, repeatable-looking output.
- Temperature and top_p are alternative sampling controls; normally adjust only one.
Common mistakes
- Avoid changing temperature and top_p together; tune one sampling control at a time.
Example: Use a low temperature for focused completions
from openai import OpenAI
client = OpenAI()
response = client.completions.create(
model="gpt-3.5-turbo-instruct",
prompt="Write a haiku about winter.",
temperature=0.2,
max_tokens=30,
)
print(response.choices[0].text)This runnable Python example requests a focused completion by setting temperature to 0.2.
When to use it
Creative generation
Increase creative variation in prose or brainstorming output.
Sampling control
Tune output randomness while leaving top_p unchanged.
Related
top_pmax_tokensmodelprompt