Skip to main content
Updated: 2026-09-10 OpenAI released ChatGPT Images 2.5 on 2026-09-08 with two API models: GPT Image 2.5 Flare and GPT Image 2.5 Sunburst. Both are listed in the BetterToken model plaza. Check it for current availability and pricing.

​
Choose a model

Both accept text and image inputs and produce images. Start with Flare for everyday tasks; compare Sunburst when you need repeated, precise edits to the same asset. Use the full model ID in the table, not just gpt-image-2.5.

​
What changed

  • Faster generation: OpenAI reports 50% lower generation latency for Flare than GPT Image 2. Actual time also depends on size, quality, and service load.
  • More precise editing: Better preservation of reference subjects, composition, and details you did not ask to change, including across multiple edits.
  • Visual and layout improvements: Better lighting, textures, style adherence, and handling of complex visual instructions. Review text and factual details before using a finished image.

​
New options in the official API

OpenAI’s image API offers these options for both models:
  • quality accepts low, medium, high, xhigh, max, and auto.
  • For a transparent background, use background: "transparent" with output_format: "png" or "webp".
  • Custom dimensions are supported. Width and height must be multiples of 16, the aspect ratio must be between 1:3 and 3:1, neither edge may exceed 3840 pixels, and total pixels must be between 655,360 and 8,294,400. Output above 2560x1440 is experimental.
These are OpenAI capabilities. For parameters supported by BetterToken, follow the image generation and image editing documentation. Sketch, templates, and image comments in ChatGPT are client features, not image API parameters.

​
Connect through BetterToken

Create a BetterToken API Key with access to the model and use these settings:

​
Generate an image

This example uses Flare. To use Sunburst, replace model with gpt-image-2.5-sunburst.
The image endpoint waits for generation to finish, then returns image content in data[0].b64_json. Check that the request succeeded before decoding and saving it. See image generation for a complete saving example. To edit an image, send multipart/form-data to POST /v1/images/edits, use the same full model ID, and upload a reference with image. See image editing for examples.

​
Integration tips

  • Start with n: 1, output_format: "png", and response_format: "b64_json" before adjusting other parameters.
  • Allow several minutes for image generation. After changing models, compare actual latency, output dimensions, and editing results.
  • OpenAI charges by tokens. Check the model plaza for BetterToken billing and prices instead of applying official estimates directly.

​
Official sources