GPT Image 2.5 API

Generate and edit images with an asynchronous REST API. Submit a request, keep the task ID, and poll for the final image URLs.

Quickstart

1. Sign in or register, then open API keys.
2. Create a named key and set GPTIMAGE25_API_KEY in your server environment.
3. Check your credit balance. Each generation costs 4 credits at 1K, 6 at 2K or 10 at 4K.
4. Submit one request. Persist data.task_id and poll with the same key.
5. On SUCCESS, read data.response and download the images to your storage.

export GPTIMAGE25_API_KEY='YOUR_API_KEY'

curl --fail-with-body 'https://gptimage25api.com/api/v1/images/generate' \
  -H "Authorization: Bearer $GPTIMAGE25_API_KEY" \
  -H 'Content-Type: application/json' \
  --data '{
    "model": "gpt-image-2.5",
    "variant": "flare",
    "prompt": "Studio photograph of a translucent green glass chair",
    "aspect_ratio": "3:2",
    "resolution": "1K",
    "public": false
  }'

# Use data.task_id from the response, with the SAME key.
curl --fail-with-body \
  'https://gptimage25api.com/api/v1/images/status?task_id=YOUR_TASK_ID' \
  -H "Authorization: Bearer $GPTIMAGE25_API_KEY"

Authentication

Both endpoints require Authorization: Bearer YOUR_API_KEY. Use an active key created on this site, associated with project gptimage25api-com. OpenAI keys, account login JWTs and keys from other projects are not accepted. A login cookie is not required.

Send requests from your server. Never expose keys in browser bundles, URLs, analytics, chat or source control. No public CORS support is enabled. Create, copy and delete keys in the account manager. A task can only be polled with the key that created it; finish outstanding tasks before deleting that key during rotation.

Generate image

POST https://gptimage25api.com/api/v1/images/generate

Send Content-Type: application/json. Supply a nonempty prompt (up to 20,000 characters) and optionally up to 16 reference images. This endpoint returns a task, not a completed image.

Unknown fields are stripped. This interface does not implement OpenAI SDK compatibility, streaming, webhook callbacks or a batch endpoint. Image edits use publicly accessible HTTP(S) image URLs; upload local files to your own storage first.

Accepted response

{
  "code": 200,
  "message": "success",
  "data": {
    "task_id": "n42YOUR_TASK_IDgptimg",
    "status": "IN_PROGRESS"
  }
}

Task status

GET https://gptimage25api.com/api/v1/images/status?task_id=YOUR_TASK_ID

task_id is required; taskId is an alias. Use the same API key used for generation. Poll about every 8 seconds and back off on transient errors. Keep a bounded deadline and support resuming status checks later.

Successful result

{
  "code": 200,
  "message": "success",
  "data": {
    "task_id": "n42YOUR_TASK_IDgptimg",
    "status": "SUCCESS",
    "consumed_credits": 4,
    "created_at": "2026-09-09 08:00:00",
    "error_message": null,
    "request": {
      "model": "gpt-image-2.5",
      "variant": "flare",
      "prompt": "Studio photograph of a translucent green glass chair",
      "aspect_ratio": "3:2",
      "resolution": "1K",
      "public": false
    },
    "response": [
      "https://your-image-storage.example/result.png"
    ]
  }
}

response can be null before success. On success it is an array of image URLs. request echoes the normalized input, including the selected GPT Image 2.5 variant. consumed_credits reports the task's credit value; failed or refunded tasks may show zero.

Complete integration examples

export GPTIMAGE25_API_KEY='YOUR_API_KEY'

curl --fail-with-body 'https://gptimage25api.com/api/v1/images/generate' \
  -H "Authorization: Bearer $GPTIMAGE25_API_KEY" \
  -H 'Content-Type: application/json' \
  --data '{
    "model": "gpt-image-2.5",
    "variant": "flare",
    "prompt": "Studio photograph of a translucent green glass chair",
    "aspect_ratio": "3:2",
    "resolution": "1K",
    "public": false
  }'

# Use data.task_id from the response, with the SAME key.
curl --fail-with-body \
  'https://gptimage25api.com/api/v1/images/status?task_id=YOUR_TASK_ID' \
  -H "Authorization: Bearer $GPTIMAGE25_API_KEY"

Errors and safe retries

Check both the HTTP status and the { code, message, data } envelope. Error data may be null or contain a task ID and status; retain a returned task ID even on failure. Other upstream 4xx/5xx statuses may be preserved.

Credits, storage and operations

Flare and Sunburst each cost 4 credits at 1K, 6 at 2K and 10 at 4K, for both text-to-image and image editing. Billing and refunds remain handled by the generation backend. Failed tasks may enter REFUND_PENDING; check the task and account balance for the final result.

Use public: false for private tasks. Store images in your own storage promptly; permanent result retention is not promised. There is no published concurrency quota or delivery-time SLA here. Account pricing depends on the configured payment store.

For AI-assisted integration, use llms.txt, the full guide, OpenAPI or the coding prompt. These resources share the same API contract and examples.