GPT Image 2.6
GPT Image 2.6 image inpainting API. Image inputs, Transparent background, Mask. Explore supported settings, pricing and runnable examples on Fusara.
Your result will appear here
Configure your request and select Generate.
MODEL GUIDE
Build with GPT Image 2.6
GPT Image 2.6 image inpainting API. Image inputs, Transparent background, Mask. Explore supported settings, pricing and runnable examples on Fusara.
How to use this model
- Choose inpaint an image and adjust the settings above.
- Add the input media required by your task. Use a public URL or a media ID from your account.
- Run the request, review the result, then copy the API code.
One request, any integration
Send POST /api/v3/run with model: openai-gpt-image26-inpaint/flare. The API returns the run ID and expected cost. Read GET /api/v3/run/{id} for its status and result URLs, or provide a webhook URL.
In MCP, use run with the same model and parameters.
Versions: Flare (default), Sunburst.
Parameters
These fields come from this model’s API schema. The server also validates input combinations and conditional limits.
| Parameter | Type | Requirements / values / default |
|---|---|---|
background | string | Optional · Values: "auto", "opaque", "transparent" · Default: "auto" |
prompt | string | Required · Minimum 1 character · Maximum 32000 characters |
images | array | Required · Minimum 1 item when supplied · Maximum 1 item |
webhook_url | string | Optional · Optional public HTTPS callback URL; no registration required. Receives the full Run JSON, ready to save without another GET. Signature verification is optional; without it, sender authenticity is unverified. · Maximum 2048 characters |
How do billing and results work?
Estimate the current request with POST /api/v3/run/estimate before generating. Generations use your Fusara balance. A successful run includes direct output URLs; save results you need before their links expire. An estimate does not create a run.
Can I use this model from an agent?
Connect your client to https://fusara.ai/mcp. Use get_model to discover the current schema, estimate_run for the expected cost, and run to start.