4Z API

Create gpt-image-2.6-flare

Endpoint: POST https://4zapi.com/v1/images/generations

gpt-image-2.6-flare is the fast, economical tier of the gpt-image-2.6 series: quick generation at lower cost, suited to batch previews, drafts, and fast GIF frames. For higher quality, richer detail and lighting, or consistent subjects across multi-round local edits, use gpt-image-2.6-sunburst.

  • Synchronous: the image is returned directly, no polling.
  • Parameters match gpt-image-2.5; quality supports xhigh and max.
  • Response format: b64_json when response_format is omitted; url returns an image link valid for 7 days, so save it promptly.
  • For image editing use POST /v1/images/edits with model gpt-image-2.6-flare.

Parameters

  • Content-Type (header, string, required)

Request body (application/json)

  • model (string, required): Model name gpt-image-2.6-flare fast iteration, low latency, for interactive generation and rapid drafts gpt-image-2.6-sunburst detailed visual work, steadier complex layouts and in-image text, slower ["gpt-image-2.6-flare", "gpt-image-2.6-sunburst"]
  • prompt (string, required): A text description of the desired image. Maximum length is 32,000 characters. For GIF output, describe the subject, the motion, and the desired loop.
  • background (string): Background Options: auto (default), opaque, transparent transparent requires output_format to be png or webp ["auto", "opaque", "transparent"]
  • complexity (string): Complexity (new in 2.6): computation spent planning the image, independently of rendering quality. Options: low, medium, high (omitted = model default, which is high) low suits focused edits (background swaps, signage changes); high suits dense text, complex layouts, and object relationships. Higher complexity increases latency and may introduce unwanted changes; compare results for your task. For SVG output, complexity controls the reference image, independently of svg.reasoning.effort. Adjusting complexity has no separate fee. ["low", "medium", "high"]
  • frame_count (integer): GIF frame count (2-16), only with output_format=gif. Defaults to 9. Frame count × frame size × n must fit the model's pixel budget; reduce frame_count, size, or n if the request exceeds it.
  • frame_duration_ms (integer): Delay per GIF frame in milliseconds, only with output_format=gif. Defaults to 200, minimum 20. Rounded to the nearest 10 ms; the encoded delay is capped at 655,350 ms per frame.
  • moderation (string): Content moderation strictness Options: auto (default), low ["auto", "low"]
  • n (integer): The number of images to generate. Must be between 1 and 10. Must be 1 for SVG output.
  • output_compression (integer): Compression level (0-100), for jpeg / webp only. Defaults to 100.
  • output_format (string): Image format Options: png (default), jpeg, webp, gif, svg Use png or webp for a transparent background. gif is the new animated output (pair with frame_count / frame_duration_ms); svg is the new vector output, generations only, and requires n=1. gif and svg do not support streaming, partial_images, prompt_rewriter, or custom output_compression. (Official field name is output_format; earlier docs written as format refer to this field.) ["png", "jpeg", "webp", "gif", "svg"]
  • quality (string): Image quality Options: low, medium, high, xhigh, max, auto (default) low is for drafts; xhigh / max are finer but slower and more expensive. For SVG output, quality controls the rendering quality of the reference image. ["low", "medium", "high", "xhigh", "max", "auto"]
  • response_format (string): url: return an image URL b64_json: return Base64 image data ["url", "b64_json"]
  • size (string): Image size 1024x1024 square 1536x1024 landscape 1024x1536 portrait auto default, chosen by the model Custom dimensions as WIDTHxHEIGHT (e.g. 1536x864): 1. Both width and height must be multiples of 16px 2. Aspect ratio between 1:3 and 3:1 3. Max edge 3840px 4. Total pixels 655,360 to 8,294,400 Resolutions above 2560x1440 are experimental. For GIF output, size describes each frame; frame count × frame size must fit the model's pixel budget.
  • svg (object): SVG author model and reasoning settings (only with output_format=svg). The top-level model selects the image model that creates the reference image; svg.model selects the author model. When model selection is enabled, supply svg.model — omitting it raises: SVG generation requires '***.model'. If you see that error, the svg object was not sent (or not passed through). Billing = reference image at the image model rate + SVG tokens at the author model rate; the response's svg_usage reports the author model usage.

Request example

{
  "model": "gpt-image-2.6-flare",
  "n": 1,
  "prompt": "一只橘猫趴在窗台上晒太阳，水彩画风格",
  "size": "1024x1024"
}