Structured Outputs

Structured Outputs#

Ensure responses adhere to a JSON schema.

Introduction#

JSON is one of the most widely used formats in the world for applications to exchange data.

Structured Outputs is a feature that ensures the model will always generate responses that adhere to your supplied JSON Schema, so you don’t need to worry about the model omitting a required key, or hallucinating an invalid enum value.

Some benefits of Structured Outputs include:

Reliable type-safety: No need to validate or retry incorrectly formatted responses
Explicit refusals: Safety-based model refusals are now programmatically detectable
Simpler prompting: No need for strongly worded prompts to achieve consistent formatting

In addition to supporting JSON Schema in the REST API, the OpenAI SDKs for Python and JavaScript also make it easy to define object schemas using Pydantic and Zod respectively. Below, you can see how to extract information from unstructured text that conforms to a schema defined in code.

Getting a structured response

from openai import OpenAI
from pydantic import BaseModel

client = OpenAI()

class CalendarEvent(BaseModel):
    name: str
    date: str
    participants: list[str]

response = client.responses.parse(
    model="gpt-4o-2024-08-06",
    input=[
        {"role": "system", "content": "Extract the event information."},
        {
            "role": "user",
            "content": "Alice and Bob are going to a science fair on Friday.",
        },
    ],
    text_format=CalendarEvent,
)

event = response.output_parsed

Supported models#

Structured Outputs is available in our latest large language models, starting with GPT-4o. Older models like gpt-4-turbo and earlier may use JSON mode instead.

When to use Structured Outputs via function calling vs via text.format#

Structured Outputs is available in two forms in the OpenAI API:

When using function calling
When using a json_schema response format

Function calling is useful when you are building an application that bridges the models and functionality of your application.

For example, you can give the model access to functions that query a database in order to build an AI assistant that can help users with their orders, or functions that can interact with the UI.

Conversely, Structured Outputs via response_format are more suitable when you want to indicate a structured schema for use when the model responds to the user, rather than when the model calls a tool.

For example, if you are building a math tutoring application, you might want the assistant to respond to your user using a specific JSON Schema so that you can generate a UI that displays different parts of the model’s output in distinct ways.

Put simply:

If you are connecting the model to tools, functions, data, etc. in your system, then you should use function calling
If you want to structure the model’s output when it responds to the user, then you should use a structured text.format

Structured Outputs vs JSON mode#

Structured Outputs is the evolution of JSON mode. While both ensure valid JSON is produced, only Structured Outputs ensure schema adherance. Both Structured Outputs and JSON mode are supported in the Responses API,Chat Completions API, Assistants API, Fine-tuning API and Batch API.

We recommend always using Structured Outputs instead of JSON mode when possible.

However, Structured Outputs with response_format: {type: "json_schema", ...} is only supported with the gpt-4o-mini, gpt-4o-mini-2024-07-18, and gpt-4o-2024-08-06 model snapshots and later.

	Structured Outputs	JSON Mode
Outputs valid JSON	Yes	Yes
Adheres to schema	Yes (see supported schemas)	No
Compatible models	gpt-4o-mini, gpt-4o-2024-08-06, and later	gpt-3.5-turbo, gpt-4-* and gpt-4o-* models
Enabling	text: { format: { type: “json_schema”, “strict”: true, “schema”: … } }	text: { format: { type: “json_object” } }

Examples#

Chain of thought

Chain of thought#

You can ask the model to output an answer in a structured, step-by-step way, to guide the user through the solution.

Structured Outputs for chain-of-thought math tutoring

from openai import OpenAI
from pydantic import BaseModel

client = OpenAI()

class Step(BaseModel):
    explanation: str
    output: str

class MathReasoning(BaseModel):
    steps: list[Step]
    final_answer: str

response = client.responses.parse(
    model="gpt-4o-2024-08-06",
    input=[
        {
            "role": "system",
            "content": "You are a helpful math tutor. Guide the user through the solution step by step.",
        },
        {"role": "user", "content": "how can I solve 8x + 7 = -23"},
    ],
    text_format=MathReasoning,
)

math_reasoning = response.output_parsed

Example response#

{
  "steps": [
    {
      "explanation": "Start with the equation 8x + 7 = -23.",
      "output": "8x + 7 = -23"
    },
    {
      "explanation": "Subtract 7 from both sides to isolate the term with the variable.",
      "output": "8x = -23 - 7"
    },
    {
      "explanation": "Simplify the right side of the equation.",
      "output": "8x = -30"
    },
    {
      "explanation": "Divide both sides by 8 to solve for x.",
      "output": "x = -30 / 8"
    },
    {
      "explanation": "Simplify the fraction.",
      "output": "x = -15 / 4"
    }
  ],
  "final_answer": "x = -15 / 4"
}

Structured data extraction#

You can define structured fields to extract from unstructured input data, such as research papers.

Extracting data from research papers using Structured Outputs

from openai import OpenAI
from pydantic import BaseModel

client = OpenAI()

class ResearchPaperExtraction(BaseModel):
    title: str
    authors: list[str]
    abstract: str
    keywords: list[str]

response = client.responses.parse(
    model="gpt-4o-2024-08-06",
    input=[
        {
            "role": "system",
            "content": "You are an expert at structured data extraction. You will be given unstructured text from a research paper and should convert it into the given structure.",
        },
        {"role": "user", "content": "..."},
    ],
    text_format=ResearchPaperExtraction,
)

research_paper = response.output_parsed

Example response#

{
  "title": "Application of Quantum Algorithms in Interstellar Navigation: A New Frontier",
  "authors": [
    "Dr. Stella Voyager",
    "Dr. Nova Star",
    "Dr. Lyra Hunter"
  ],
  "abstract": "This paper investigates the utilization of quantum algorithms to improve interstellar navigation systems. By leveraging quantum superposition and entanglement, our proposed navigation system can calculate optimal travel paths through space-time anomalies more efficiently than classical methods. Experimental simulations suggest a significant reduction in travel time and fuel consumption for interstellar missions.",
  "keywords": [
    "Quantum algorithms",
    "interstellar navigation",
    "space-time anomalies",
    "quantum superposition",
    "quantum entanglement",
    "space travel"
  ]
}

UI Generation#

You can generate valid HTML by representing it as recursive data structures with constraints, like enums.

Generating HTML using Structured Outputs

from enum import Enum
from typing import List

from openai import OpenAI
from pydantic import BaseModel

client = OpenAI()

class UIType(str, Enum):
    div = "div"
    button = "button"
    header = "header"
    section = "section"
    field = "field"
    form = "form"

class Attribute(BaseModel):
    name: str
    value: str

class UI(BaseModel):
    type: UIType
    label: str
    children: List["UI"]
    attributes: List[Attribute]

UI.model_rebuild()  # This is required to enable recursive types

class Response(BaseModel):
    ui: UI

response = client.responses.parse(
    model="gpt-4o-2024-08-06",
    input=[
        {
            "role": "system",
            "content": "You are a UI generator AI. Convert the user input into a UI.",
        },
        {"role": "user", "content": "Make a User Profile Form"},
    ],
    text_format=Response,
)

ui = response.output_parsed

Example response#

{
  "type": "form",
  "label": "User Profile Form",
  "children": [
    {
      "type": "div",
      "label": "",
      "children": [
        {
          "type": "field",
          "label": "First Name",
          "children": [],
          "attributes": [
            {
              "name": "type",
              "value": "text"
            },
            {
              "name": "name",
              "value": "firstName"
            },
            {
              "name": "placeholder",
              "value": "Enter your first name"
            }
          ]
        },
        {
          "type": "field",
          "label": "Last Name",
          "children": [],
          "attributes": [
            {
              "name": "type",
              "value": "text"
            },
            {
              "name": "name",
              "value": "lastName"
            },
            {
              "name": "placeholder",
              "value": "Enter your last name"
            }
          ]
        }
      ],
      "attributes": []
    },
    {
      "type": "button",
      "label": "Submit",
      "children": [],
      "attributes": [
        {
          "name": "type",
          "value": "submit"
        }
      ]
    }
  ],
  "attributes": [
    {
      "name": "method",
      "value": "post"
    },
    {
      "name": "action",
      "value": "/submit-profile"
    }
  ]
}

Moderation#

You can classify inputs on multiple categories, which is a common way of doing moderation.

Moderation using Structured Outputs

from enum import Enum
from typing import Optional

from openai import OpenAI
from pydantic import BaseModel

client = OpenAI()

class Category(str, Enum):
    violence = "violence"
    sexual = "sexual"
    self_harm = "self_harm"

class ContentCompliance(BaseModel):
    is_violating: bool
    category: Optional[Category]
    explanation_if_violating: Optional[str]

response = client.responses.parse(
    model="gpt-4o-2024-08-06",
    input=[
        {
            "role": "system",
            "content": "Determine if the user input violates specific guidelines and explain if they do.",
        },
        {"role": "user", "content": "How do I prepare for a job interview?"},
    ],
    text_format=ContentCompliance,
)

compliance = response.output_parsed

Example response#

{
  "is_violating": false,
  "category": null,
  "explanation_if_violating": null
}

How to use Structured Outputs with text.format#

Step 1: Define your schema#

First you must design the JSON Schema that the model should be constrained to follow. See the examples at the top of this guide for reference.

While Structured Outputs supports much of JSON Schema, some features are unavailable either for performance or technical reasons. See here for more details.

Tips for your JSON Schema#

To maximize the quality of model generations, we recommend the following:

Name keys clearly and intuitively
Create clear titles and descriptions for important keys in your structure
Create and use evals to determine the structure that works best for your use case

Step 2: Supply your schema in the API call#

To use Structured Outputs, simply specify

text: { format: { type: "json_schema", "strict": true, "schema": … } }

For example:

response = client.responses.create(
    model="gpt-4o-2024-08-06",
    input=[
        {"role": "system", "content": "You are a helpful math tutor. Guide the user through the solution step by step."},
        {"role": "user", "content": "how can I solve 8x + 7 = -23"}
    ],
    text={
        "format": {
            "type": "json_schema",
            "name": "math_response",
            "schema": {
                "type": "object",
                "properties": {
                    "steps": {
                        "type": "array",
                        "items": {
                            "type": "object",
                            "properties": {
                                "explanation": {"type": "string"},
                                "output": {"type": "string"}
                            },
                            "required": ["explanation", "output"],
                            "additionalProperties": False
                        }
                    },
                    "final_answer": {"type": "string"}
                },
                "required": ["steps", "final_answer"],
                "additionalProperties": False
            },
            "strict": True
        }
    }
)

print(response.output_text)

Note: the first request you make with any schema will have additional latency as our API processes the schema, but subsequent requests with the same schema will not have additional latency.

Step 3: Handle edge cases#

In some cases, the model might not generate a valid response that matches the provided JSON schema.

This can happen in the case of a refusal, if the model refuses to answer for safety reasons, or if for example you reach a max tokens limit and the response is incomplete.

try:
    response = client.responses.create(
        model="gpt-4o-2024-08-06",
        input=[
            {
                "role": "system",
                "content": "You are a helpful math tutor. Guide the user through the solution step by step.",
            },
            {"role": "user", "content": "how can I solve 8x + 7 = -23"},
        ],
        text={
            "format": {
                "type": "json_schema",
                "name": "math_response",
                "strict": True,
                "schema": {
                    "type": "object",
                    "properties": {
                        "steps": {
                            "type": "array",
                            "items": {
                                "type": "object",
                                "properties": {
                                    "explanation": {"type": "string"},
                                    "output": {"type": "string"},
                                },
                                "required": ["explanation", "output"],
                                "additionalProperties": False,
                            },
                        },
                        "final_answer": {"type": "string"},
                    },
                    "required": ["steps", "final_answer"],
                    "additionalProperties": False,
                },
                "strict": True,
            },
        },
    )
except Exception as e:
    # handle errors like finish_reason, refusal, content_filter, etc.
    pass

Refusals with Structured Outputs#

When using Structured Outputs with user-generated input, OpenAI models may occasionally refuse to fulfill the request for safety reasons. Since a refusal does not necessarily follow the schema you have supplied in response_format, the API response will include a new field called refusal to indicate that the model refused to fulfill the request.

When the refusal property appears in your output object, you might present the refusal in your UI, or include conditional logic in code that consumes the response to handle the case of a refused request.

class Step(BaseModel):
    explanation: str
    output: str

class MathReasoning(BaseModel):
    steps: list[Step]
    final_answer: str

completion = client.beta.chat.completions.parse(
    model="gpt-4o-2024-08-06",
    messages=[
        {"role": "system", "content": "You are a helpful math tutor. Guide the user through the solution step by step."},
        {"role": "user", "content": "how can I solve 8x + 7 = -23"}
    ],
    response_format=MathReasoning,
)

math_reasoning = completion.choices[0].message

# If the model refuses to respond, you will get a refusal message
if (math_reasoning.refusal):
    print(math_reasoning.refusal)
else:
    print(math_reasoning.parsed)

The API response from a refusal will look something like this:

{
  "id": "resp_1234567890",
  "object": "response",
  "created_at": 1721596428,
  "status": "completed",
  "error": null,
  "incomplete_details": null,
  "input": [],
  "instructions": null,
  "max_output_tokens": null,
  "model": "gpt-4o-2024-08-06",
  "output": [{
    "id": "msg_1234567890",
    "type": "message",
    "role": "assistant",
    "content": [
      {
        "type": "refusal",
        "refusal": "I'm sorry, I cannot assist with that request."
      }
    ]
  }],
  "usage": {
    "input_tokens": 81,
    "output_tokens": 11,
    "total_tokens": 92,
    "output_tokens_details": {
      "reasoning_tokens": 0,
    }
  },
}

Tips and best practices#

Handling user-generated input#

If your application is using user-generated input, make sure your prompt includes instructions on how to handle situations where the input cannot result in a valid response.

The model will always try to adhere to the provided schema, which can result in hallucinations if the input is completely unrelated to the schema.

You could include language in your prompt to specify that you want to return empty parameters, or a specific sentence, if the model detects that the input is incompatible with the task.

Handling mistakes#

Structured Outputs can still contain mistakes. If you see mistakes, try adjusting your instructions, providing examples in the system instructions, or splitting tasks into simpler subtasks. Refer to the prompt engineering.

Avoid JSON schema divergence#

To prevent your JSON Schema and corresponding types in your programming language from diverging, we strongly recommend using the native Pydantic/zod sdk support.

If you prefer to specify the JSON schema directly, you could add CI rules that flag when either the JSON schema or underlying data objects are edited, or add a CI step that auto-generates the JSON Schema from type definitions (or vice-versa).

Streaming#

You can use streaming to process model responses or function call arguments as they are being generated, and parse them as structured data.

That way, you don’t have to wait for the entire response to complete before handling it. This is particularly useful if you would like to display JSON fields one by one, or handle function call arguments as soon as they are available.

We recommend relying on the SDKs to handle streaming with Structured Outputs.

from typing import List

from openai import OpenAI
from pydantic import BaseModel

class EntitiesModel(BaseModel):
    attributes: List[str]
    colors: List[str]
    animals: List[str]

client = OpenAI()

with client.responses.stream(
    model="gpt-4.1",
    input=[
        {"role": "system", "content": "Extract entities from the input text"},
        {
            "role": "user",
            "content": "The quick brown fox jumps over the lazy dog with piercing blue eyes",
        },
    ],
    text_format=EntitiesModel,
) as stream:
    for event in stream:
        if event.type == "response.refusal.delta":
            print(event.delta, end="")
        elif event.type == "response.output_text.delta":
            print(event.delta, end="")
        elif event.type == "response.error":
            print(event.error, end="")
        elif event.type == "response.completed":
            print("Completed")
            # print(event.response.output)

    final_response = stream.get_final_response()
    print(final_response)

Supported schemas#

Structured Outputs supports a subset of the JSON Schema language.

Supported types#

The following types are supported for Structured Outputs:

String
Number
Boolean
Integer
Object
Array
Enum
anyOf

Supported properties#

In addition to specifying the type of a property, you can specify a selection of additional constraints:

Supported string properties:

pattern — A regular expression that the string must match.
format — Predefined formats for strings. Currently supported:
- date-time
- time
- date
- duration
- email
- hostname
- ipv4
- ipv6
- uuid

Supported number properties:

multipleOf — The number must be a multiple of this value.
maximum — The number must be less than or equal to this value.
exclusiveMaximum — The number must be less than this value.
minimum — The number must be greater than or equal to this value.
exclusiveMinimum — The number must be greater than this value.

Supported array properties:

minItems — The array must have at least this many items.
maxItems — The array must have at most this many items.

JSON mode#

JSON mode is a more basic version of the Structured Outputs feature. While JSON mode ensures that model output is valid JSON, Structured Outputs reliably matches the model’s output to the schema you specify. We recommend you use Structured Outputs if it is supported for your use case.

When JSON mode is turned on, the model’s output is ensured to be valid JSON, except for in some edge cases that you should detect and handle appropriately.

To turn on JSON mode with the Responses API you can set the text.format to { "type": "json_object" }. If you are using function calling, JSON mode is always turned on.

Important notes:

When using JSON mode, you must always instruct the model to produce JSON via some message in the conversation, for example via your system message. If you don’t include an explicit instruction to generate JSON, the model may generate an unending stream of whitespace and the request may run continually until it reaches the token limit. To help ensure you don’t forget, the API will throw an error if the string “JSON” does not appear somewhere in the context.
JSON mode will not guarantee the output matches any specific schema, only that it is valid and parses without errors. You should use Structured Outputs to ensure it matches your schema, or if that is not possible, you should use a validation library and potentially retries to ensure that the output matches your desired schema.
Your application must detect and handle the edge cases that can result in the model output not being a complete JSON object (see below)

Handling edge cases

we_did_not_specify_stop_tokens = True

try:
    response = client.responses.create(
        model="gpt-3.5-turbo-0125",
        input=[
            {"role": "system", "content": "You are a helpful assistant designed to output JSON."},
            {"role": "user", "content": "Who won the world series in 2020? Please respond in the format {winner: ...}"}
        ],
        text={"format": {"type": "json_object"}}
    )

    # Check if the conversation was too long for the context window, resulting in incomplete JSON 
    if response.status == "incomplete" and response.incomplete_details.reason == "max_output_tokens":
        # your code should handle this error case
        pass

    # Check if the OpenAI safety system refused the request and generated a refusal instead
    if response.output[0].content[0].type == "refusal":
        # your code should handle this error case
        # In this case, the .content field will contain the explanation (if any) that the model generated for why it is refusing
        print(response.output[0].content[0]["refusal"])

    # Check if the model's output included restricted content, so the generation of JSON was halted and may be partial
    if response.status == "incomplete" and response.incomplete_details.reason == "content_filter":
        # your code should handle this error case
        pass

    if response.status == "completed":
        # In this case the model has either successfully finished generating the JSON object according to your schema, or the model generated one of the tokens you provided as a "stop token"

        if we_did_not_specify_stop_tokens:
            # If you didn't specify any stop tokens, then the generation is complete and the content key will contain the serialized JSON object
            # This will parse successfully and should now contain  "{"winner": "Los Angeles Dodgers"}"
            print(response.output_text)
        else:
            # Check if the response.output_text ends with one of your stop tokens and handle appropriately
            pass
except Exception as e:
    # Your code should handle errors here, for example a network error calling the API
    print(e)

Resources#

To learn more about Structured Outputs, we recommend browsing the following resources:

Check out our introductory cookbook on Structured Outputs

Structured Outputs

Contents

Structured Outputs#

Introduction#

Supported models#

When to use Structured Outputs via function calling vs via text.format#

Structured Outputs vs JSON mode#

Examples#

Chain of thought#

Example response#

Structured data extraction#

Example response#

UI Generation#

Example response#

Moderation#

Example response#

How to use Structured Outputs with text.format#

Step 1: Define your schema#

Tips for your JSON Schema#

Step 2: Supply your schema in the API call#

Step 3: Handle edge cases#

Refusals with Structured Outputs#

Tips and best practices#

Handling user-generated input#

Handling mistakes#

Avoid JSON schema divergence#

Streaming#

Supported schemas#

Supported types#

Supported properties#

JSON mode#

Resources#