GPT Image 1 API: Shutdown Status and Migration to GPT Image 2.6

As of November 10, 2026, gpt-image-1 is no longer available through the direct OpenAI API. OpenAI lists it as shut down on October 23, 2026. For a new integration or migration, use a currently supported model such as gpt-image-2.6-flare or gpt-image-2.6-sunburst. OpenAI describes Flare as the faster-iteration choice and Sunburst as the option for more detailed work or precise edits. (OpenAI API model reference, OpenAI image prompting guide)

Call the direct OpenAI Images API

For a single image generated from a prompt, use the Images API. Its generation operation is POST /v1/images/generations; the direct OpenAI endpoint is https://api.openai.com/v1/images/generations. Send an Authorization: Bearer $OPENAI_API_KEY header and Content-Type: application/json. The model goes in the JSON body. Do not substitute an Azure resource endpoint or a third-party gateway URL. (OpenAI Images API reference, OpenAI API authentication reference)

Before you run the example

• Create an OpenAI API key and make it available to your server-side application as OPENAI_API_KEY.
• Install the Python SDK with pip install --upgrade openai.
• Check model access: OpenAI says you may need to complete API Organization Verification in the developer console before using GPT Image models. (Image generation guide)
• Keep the key on your server; do not embed it in browser or mobile-app code. (API authentication reference)

On macOS or Linux, set the key in your shell session:

export OPENAI_API_KEY="your_api_key"

Python: generate and save a PNG

This example explicitly requests PNG output, then decodes the returned image data and writes generated.png in the current directory. The API reference documents b64_json as the generated-image field returned by default for GPT Image models; 1024x1024 and medium are documented size and quality options for GPT Image 2.6. (Images API reference, image prompting guide)

import base64
from openai import OpenAI

client = OpenAI()  # Reads OPENAI_API_KEY from the environment

result = client.images.generate(
    model="gpt-image-2.6-flare",
    prompt="A small glass greenhouse in a garden at sunrise, editorial illustration",
    size="1024x1024",
    quality="medium",
    output_format="png",
)

image_bytes = base64.b64decode(result.data[0].b64_json)
with open("generated.png", "wb") as image_file:
    image_file.write(image_bytes)

print("Saved generated.png")

To try Sunburst instead, change model to gpt-image-2.6-sunburst. For a different output type, check the current Image API options: the guide lists PNG, JPEG, WebP, SVG, and GIF generation. (Image prompting guide)

Migrating an existing gpt-image-1 integration

1. Replace the retired model ID. Select a supported model, such as gpt-image-2.6-flare or gpt-image-2.6-sunburst; do not continue sending requests to gpt-image-1. (Images API reference)
2. Check request options and output handling. Explicitly set the output format if your application expects a particular file type. Decode b64_json for GPT Image output rather than assuming the response contains an image URL. (Images API reference)
3. Test before switching production traffic. Compare your existing prompts and reference images, output dimensions, edit behavior, latency, and downstream file handling. Treat this as a migration test, not a guarantee that different models produce identical results.
4. Protect the production credential. Keep it in server-side environment configuration or a secrets manager. For an application workload, OpenAI documents project service accounts as a separate identity option from a developer’s personal key. (API keys and service accounts)

Access and common request failures

GPT Image use may require API Organization Verification; check the notice in your developer console if a request reports an access restriction. If you receive a 401, check that the key is correct and that the request is using the intended organization. (Image generation guide, OpenAI error codes)

For a 429, read the response error details before retrying. OpenAI’s guidance distinguishes rate-limit responses from exhausted credits or organization/project spending limits. Check the applicable model and current limits in your organization’s Limits page; pace requests and use retry guidance for rate limits, but address the balance or spending limit when the error identifies billing as the cause. (OpenAI rate-limit troubleshooting)

Direct OpenAI, Azure, and gateways are different

This walkthrough uses the direct OpenAI API. Azure OpenAI uses Azure resource endpoints, deployments, and Azure authentication; a third-party gateway uses its own endpoint and credentials. Keep each provider’s model identifier, URL, authentication, and response handling paired with that provider’s documentation.

Sources

Create image — OpenAI API Reference

API Overview — OpenAI API Reference

Image generation — OpenAI API

Image prompting — OpenAI API

Error codes — OpenAI API

Troubleshooting API rate limits and 429 errors — OpenAI Help Center

API keys and service accounts — OpenAI API