Skip to content
For the complete documentation index, see llms.txt. Markdown versions of documentation pages are available by appending .md to the page URL.
Primary navigation

Suggested
Overview Models Agents Tools Audio & voice Production API reference

Create a decision

POST/decisions

Evaluate ordered classification and scoring questions against shared input. Answers are returned in question order.

Supply input as a string or user messages containing text and inline images. Only user messages with input_text and input_image parts are supported; non-user roles, function calls, files, audio, and item references are not supported. Images require a data URL, not an external URL or file ID. At most 128 images are allowed across the request.

Each question can return a refusal instead of a scored answer. A refusal has type refusal and the corresponding question name, or null if unnamed.

Body ParametersJSONExpand Collapse
input: string or array of DecisionInputMessage { content, role, type }

Shared evidence, as a string or an array of user messages containing text and inline images. Non-user roles, function calls, function-call outputs, files, audio, and item references are not supported. At most 128 image parts are allowed across all messages in one request.

One of the following:
string
array of DecisionInputMessage { content, role, type }
content: string or array of DecisionInputPart

Text evidence or an ordered list of text and inline image parts.

One of the following:
string
Parts = array of DecisionInputPart
One of the following:
DecisionInputText object { text, type }
text: string
minLength0
maxLength10485760
type: "input_text"
DecisionInputImage object { image_url, type, detail }

An inline image. External URLs and file IDs are not supported.

image_url: string

A base64-encoded image in a data URL.

minLength0
maxLength1073741824
type: "input_image"
detail: optional "low" or "high" or "auto" or "original" or null

The image detail level, using the selected model’s image profile. Defaults to auto.

One of the following:
"low"
"high"
"auto"
"original"
role: "user"
type: optional "message"
model: string
minLength0
maxLength1048576
questions: array of object { instructions, type, name } or object { choices, instructions, type, name } or object { instructions, levels, type, name }
One of the following:
Predicate object { instructions, type, name }
instructions: string
minLength0
maxLength1048576
type: "predicate"

The type of the object. Always predicate.

name: optional string
minLength0
maxLength1048576
Choice object { choices, instructions, type, name }
choices: array of object { value, description }
value: string or boolean

Choice values are typed: a string and a boolean with the same text are distinct.

One of the following:
string
boolean
description: optional string
minLength0
maxLength1048576
instructions: string
minLength0
maxLength1048576
type: "choice"

The type of the object. Always choice.

name: optional string
minLength0
maxLength1048576
Score object { instructions, levels, type, name }
instructions: string
minLength0
maxLength1048576
levels: array of object { label, description }
label: string
minLength0
maxLength1048576
description: optional string
minLength0
maxLength1048576
type: "score"

The type of the object. Always score.

name: optional string
minLength0
maxLength1048576
safety_identifier: optional string or null

Opaque caller-provided end-user identifier, scoped by the verified org. Match Responses’ limit; this is never the authenticated user identity.

minLength0
maxLength128
ReturnsExpand Collapse
Decision object { answers, model, usage }
answers: array of object { name, probability, type } or object { choice, confidence, name, 2 more } or object { confidence, name, probabilities, 2 more } or object { name, type }
One of the following:
Predicate object { name, probability, type }
name: string or null
minLength0
probability: number
type: "predicate"

The type of the object. Always predicate.

Choice object { choice, confidence, name, 2 more }
choice: string or boolean

Choice values are typed: a string and a boolean with the same text are distinct.

One of the following:
string
boolean
confidence: number
name: string or null
minLength0
probabilities: array of object { probability, value }
probability: number
value: string or boolean

Choice values are typed: a string and a boolean with the same text are distinct.

One of the following:
string
boolean
type: "choice"

The type of the object. Always choice.

Score object { confidence, name, probabilities, 2 more }
confidence: number
name: string or null
minLength0
probabilities: array of object { label, probability, value }
label: string
minLength0
probability: number
value: number
formatint64
score: number
type: "score"

The type of the object. Always score.

Refusal object { name, type }

The host may decline one question without disclosing its refusal score.

name: string or null
minLength0
type: "refusal"

The type of the object. Always refusal.

model: string
minLength0
usage: object { input_tokens, input_tokens_details, output_tokens, 2 more }
input_tokens: number
formatint64
minimum-2147483648
maximum2147483647
input_tokens_details: object { cache_write_tokens, cached_tokens }
cache_write_tokens: number
formatint64
minimum-2147483648
maximum2147483647
cached_tokens: number
formatint64
minimum-2147483648
maximum2147483647
output_tokens: number
formatint64
minimum-2147483648
maximum2147483647
output_tokens_details: object { reasoning_tokens }
reasoning_tokens: number
formatint64
minimum-2147483648
maximum2147483647
total_tokens: number
formatint64
minimum-2147483648
maximum2147483647

Create a decision

curl https://api.openai.com/v1/decisions \
  -H "Content-Type: application/json" \
  -H "Authorization: Bearer $OPENAI_API_KEY" \
  -d '{
    "model": "gpt-6-luna",
    "input": "The package arrived with a broken screen.",
    "questions": [
      {
        "type": "predicate",
        "name": "damaged",
        "instructions": "Does the customer report a damaged item?"
      }
    ]
  }'

{
  "model": "gpt-6-luna",
  "answers": [
    {"type": "predicate", "name": "damaged", "probability": 0.95}
  ],
  "usage": {
    "input_tokens": 42,
    "input_tokens_details": {"cached_tokens": 0, "cache_write_tokens": 0},
    "output_tokens": 0,
    "output_tokens_details": {"reasoning_tokens": 0},
    "total_tokens": 42
  }
}
Returns Examples
{
  "model": "gpt-6-luna",
  "answers": [
    {"type": "predicate", "name": "damaged", "probability": 0.95}
  ],
  "usage": {
    "input_tokens": 42,
    "input_tokens_details": {"cached_tokens": 0, "cache_write_tokens": 0},
    "output_tokens": 0,
    "output_tokens_details": {"reasoning_tokens": 0},
    "total_tokens": 42
  }
}