GPT Image 2

Model ID

Use with the Workers AI binding, the AI Gateway REST API, or the Unified API.

Calling the provider directly? Use the provider's model name instead, including with provider-native endpoints or the provider's own SDK. For example, Anthropic's name for anthropic/claude-opus-4.8 is claude-opus-4-8.

Third-party
Zero data retention

OpenAI's next-generation image model that creates and edits images from text prompts, with support for multiple quality levels, sizes, and output formats. Note: transparent backgrounds are not supported — use openai/gpt-image-1.5 for transparent PNGs.

Model Info
Terms and License	link ↗
More information	link ↗
Zero data retention	Yes
Pricing	Input (per 1M tokens)$5.00 Input image (per 1M tokens)$8.00 Cached input (per 1M tokens)$1.25 Cached input image (per 1M tokens)$2.00 Output image (per 1M tokens)$30.00 Output (per 1M tokens)$10.00

Usage

const response = await env.AI.run(
  'openai/gpt-image-2',
  { prompt: 'A golden retriever puppy playing in autumn leaves' },
)
console.log(response)

curl https://api.cloudflare.com/client/v4/accounts/$CLOUDFLARE_ACCOUNT_ID/ai/run \
  --header "Authorization: Bearer $CLOUDFLARE_API_TOKEN" \
  --header "Content-Type: application/json" \
  --data '{
  "model": "openai/gpt-image-2",
  "input": {
    "prompt": "A golden retriever puppy playing in autumn leaves"
  }
}'

{
  "gatewayMetadata": {
    "keySource": "BYOK"
  },
  "result": {
    "image": "https://pub-04a6d208d361438ea01b797e6973bd19.r2.dev/catalog/openai__gpt-image-2/simple-generation.png"
  },
  "state": "Completed"
}

Examples

High Quality — Generate a high-quality detailed image

const response = await env.AI.run(
  'openai/gpt-image-2',
  {
    prompt:
      'A detailed botanical illustration of exotic tropical flowers with fine line work and watercolor textures',
    quality: 'high',
  },
)
console.log(response)

curl https://api.cloudflare.com/client/v4/accounts/$CLOUDFLARE_ACCOUNT_ID/ai/run \
  --header "Authorization: Bearer $CLOUDFLARE_API_TOKEN" \
  --header "Content-Type: application/json" \
  --data '{
  "model": "openai/gpt-image-2",
  "input": {
    "prompt": "A detailed botanical illustration of exotic tropical flowers with fine line work and watercolor textures",
    "quality": "high"
  }
}'

{
  "gatewayMetadata": {
    "keySource": "BYOK"
  },
  "result": {
    "image": "https://pub-04a6d208d361438ea01b797e6973bd19.r2.dev/catalog/openai__gpt-image-2/high-quality.png"
  },
  "state": "Completed"
}

Custom Size — Generate a portrait-oriented image

const response = await env.AI.run(
  'openai/gpt-image-2',
  {
    prompt: 'A towering redwood forest with sunbeams filtering through the canopy, misty atmosphere',
    size: '1024x1536',
  },
)
console.log(response)

curl https://api.cloudflare.com/client/v4/accounts/$CLOUDFLARE_ACCOUNT_ID/ai/run \
  --header "Authorization: Bearer $CLOUDFLARE_API_TOKEN" \
  --header "Content-Type: application/json" \
  --data '{
  "model": "openai/gpt-image-2",
  "input": {
    "prompt": "A towering redwood forest with sunbeams filtering through the canopy, misty atmosphere",
    "size": "1024x1536"
  }
}'

{
  "gatewayMetadata": {
    "keySource": "BYOK"
  },
  "result": {
    "image": "https://pub-04a6d208d361438ea01b797e6973bd19.r2.dev/catalog/openai__gpt-image-2/custom-size.png"
  },
  "state": "Completed"
}

WebP Output — Generate an image in WebP format for smaller file size

const response = await env.AI.run(
  'openai/gpt-image-2',
  {
    prompt:
      'A neon-lit cyberpunk cityscape at night with rain-slicked streets and holographic billboards',
    output_format: 'webp',
    quality: 'high',
  },
)
console.log(response)

curl https://api.cloudflare.com/client/v4/accounts/$CLOUDFLARE_ACCOUNT_ID/ai/run \
  --header "Authorization: Bearer $CLOUDFLARE_API_TOKEN" \
  --header "Content-Type: application/json" \
  --data '{
  "model": "openai/gpt-image-2",
  "input": {
    "prompt": "A neon-lit cyberpunk cityscape at night with rain-slicked streets and holographic billboards",
    "output_format": "webp",
    "quality": "high"
  }
}'

{
  "gatewayMetadata": {
    "keySource": "BYOK"
  },
  "result": {
    "image": "https://pub-04a6d208d361438ea01b797e6973bd19.r2.dev/catalog/openai__gpt-image-2/webp-output.webp"
  },
  "state": "Completed"
}

Image Edit — Edit an existing image by providing it in the images array as base64 (a raw string or a data:image/{png|jpeg|webp};base64,... URI). This routes the call to OpenAI's /v1/images/edits endpoint. The example uses a tiny 32x32 smiley-face PNG - real inputs are the full base64 encoding of your source image.

const response = await env.AI.run(
  'openai/gpt-image-2',
  {
    prompt:
      'Transform this cartoon smiley into a photorealistic 3D clay sculpture sitting on a marble pedestal, studio lighting',
    images: [
      'data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAACAAAAAgCAYAAABzenr0AAAAnklEQVR42u2XQRLAIAgD8/839i/26qFCACm0ozPe1KwcQsAoXvgcAABxpwFowl4QWITHxW0LCBhxVngF4gKIirMQyBRnIJAtrkE8AuwWnyFEgKzfS1UA+3sWTju3BGAu7gKYIfBW+Q/AAQgBeMCkt1wVsLZjcwUYG2Z9wGLHZitWk1DEisubUYt2XB5IWkSyFqG0RSxvMZi0Gc1+Ox3fm00ZJ5mGVtkAAAAASUVORK5CYII=',
    ],
  },
)
console.log(response)

curl https://api.cloudflare.com/client/v4/accounts/$CLOUDFLARE_ACCOUNT_ID/ai/run \
  --header "Authorization: Bearer $CLOUDFLARE_API_TOKEN" \
  --header "Content-Type: application/json" \
  --data '{
  "model": "openai/gpt-image-2",
  "input": {
    "prompt": "Transform this cartoon smiley into a photorealistic 3D clay sculpture sitting on a marble pedestal, studio lighting",
    "images": [
      "data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAACAAAAAgCAYAAABzenr0AAAAnklEQVR42u2XQRLAIAgD8/839i/26qFCACm0ozPe1KwcQsAoXvgcAABxpwFowl4QWITHxW0LCBhxVngF4gKIirMQyBRnIJAtrkE8AuwWnyFEgKzfS1UA+3sWTju3BGAu7gKYIfBW+Q/AAQgBeMCkt1wVsLZjcwUYG2Z9wGLHZitWk1DEisubUYt2XB5IWkSyFqG0RSxvMZi0Gc1+Ox3fm00ZJ5mGVtkAAAAASUVORK5CYII="
    ]
  }
}'

{
  "gatewayMetadata": {
    "keySource": "BYOK"
  },
  "result": {
    "image": "https://pub-04a6d208d361438ea01b797e6973bd19.r2.dev/catalog/openai__gpt-image-2/image-edit.png"
  },
  "state": "Completed"
}

Multi-Image Edit — Compose multiple input images by passing up to 16 base64 strings in the images array. The model blends the references; useful for combining subjects, styles, or reference shots. The example pairs a smiley-face PNG with a red ball PNG.

const response = await env.AI.run(
  'openai/gpt-image-2',
  {
    prompt:
      'Combine these into a single photorealistic scene: a ceramic smiley-face mug next to a red rubber ball on a sunlit wooden table',
    images: [
      'data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAACAAAAAgCAYAAABzenr0AAAAnklEQVR42u2XQRLAIAgD8/839i/26qFCACm0ozPe1KwcQsAoXvgcAABxpwFowl4QWITHxW0LCBhxVngF4gKIirMQyBRnIJAtrkE8AuwWnyFEgKzfS1UA+3sWTju3BGAu7gKYIfBW+Q/AAQgBeMCkt1wVsLZjcwUYG2Z9wGLHZitWk1DEisubUYt2XB5IWkSyFqG0RSxvMZi0Gc1+Ox3fm00ZJ5mGVtkAAAAASUVORK5CYII=',
      'data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAACAAAAAgCAYAAABzenr0AAAAhklEQVR42u2XsRHAMAgDNY73nye7OG2KgGUnGCWH76j/oTACPfnhcwIA3AoTGIFXRTADPlqjakYEDJwFWyJLAk/hrAQi4YwEouEjiVuBt+FXCVcgqntvCtjVvTUF7OremkIJlEAJ6Aikf0QSX3H6MpJYx+mBRCKSSYRSiVgucZjInGa/vY5PvB72/7IdMuAAAAAASUVORK5CYII=',
    ],
  },
)
console.log(response)

curl https://api.cloudflare.com/client/v4/accounts/$CLOUDFLARE_ACCOUNT_ID/ai/run \
  --header "Authorization: Bearer $CLOUDFLARE_API_TOKEN" \
  --header "Content-Type: application/json" \
  --data '{
  "model": "openai/gpt-image-2",
  "input": {
    "prompt": "Combine these into a single photorealistic scene: a ceramic smiley-face mug next to a red rubber ball on a sunlit wooden table",
    "images": [
      "data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAACAAAAAgCAYAAABzenr0AAAAnklEQVR42u2XQRLAIAgD8/839i/26qFCACm0ozPe1KwcQsAoXvgcAABxpwFowl4QWITHxW0LCBhxVngF4gKIirMQyBRnIJAtrkE8AuwWnyFEgKzfS1UA+3sWTju3BGAu7gKYIfBW+Q/AAQgBeMCkt1wVsLZjcwUYG2Z9wGLHZitWk1DEisubUYt2XB5IWkSyFqG0RSxvMZi0Gc1+Ox3fm00ZJ5mGVtkAAAAASUVORK5CYII=",
      "data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAACAAAAAgCAYAAABzenr0AAAAhklEQVR42u2XsRHAMAgDNY73nye7OG2KgGUnGCWH76j/oTACPfnhcwIA3AoTGIFXRTADPlqjakYEDJwFWyJLAk/hrAQi4YwEouEjiVuBt+FXCVcgqntvCtjVvTUF7OremkIJlEAJ6Aikf0QSX3H6MpJYx+mBRCKSSYRSiVgucZjInGa/vY5PvB72/7IdMuAAAAAASUVORK5CYII="
    ]
  }
}'

{
  "gatewayMetadata": {
    "keySource": "BYOK"
  },
  "result": {
    "image": "https://pub-04a6d208d361438ea01b797e6973bd19.r2.dev/catalog/openai__gpt-image-2/multi-image-edit.png"
  },
  "state": "Completed"
}

Parameters

prompt

stringrequiredText prompt describing the image to generate or edit

▶images[]

arraymaxItems: 16Input images for image editing, 1-16 entries. Each entry is base64-encoded (raw string or data:image/{png|jpeg|webp};base64,... URI).

quality

stringenum: low, medium, high, autoQuality of the generated image

size

stringpattern: ^(?:auto|\d+x\d+)$Size of the generated image: auto or WIDTHxHEIGHT. Each edge must be at most 3840 pixels, both dimensions must be multiples of 16, the aspect ratio must be between 1:3 and 3:1, and total pixels must be between 655,360 and 8,294,400.

background

stringenum: transparent, opaque, autoBackground transparency setting. Use transparent for images with no background, opaque for a solid background, or auto to let the model decide.

output_format

stringenum: png, webp, jpegOutput format for the generated image

image

stringformat: uriURL to the generated image

API Schemas (Raw)

Input

Output