GPT Image 2.5 request cheat sheet: the fields Sume accepts

ChatGPT Image 2.5 on Sume is openai/gpt-image-2.5 (Flare) and openai/gpt-image-2.5-sunburst, called through POST /v1/images. Both take text-to-image, up to 16 reference images, an optional mask_url, and background. This page collects the fields in one table, from Sume's Image API docs as read on 2026-10-02.

Auto model routing continues to use Flare, per the same docs.

Which fields can I send?

A field the model does not list is rejected with 400 unsupported_parameter. These are listed for ChatGPT Image 2.5.

What are the size rules?

Custom pixels need both edges as multiples of 16, a maximum edge of 3840, an aspect ratio of at most 3:1, and 655,360 to 8,294,400 total pixels. On edits, prefer aspect_ratio: "auto" so the output matches the reference; omitting it is not the same.

How does the response arrive?

mode defaults to sync on this route, with a wait of up to 30 seconds. A generation that outlasts it returns 202 with the job envelope, and you read the images from the standard job result endpoint. mode: "async" and "webhook" return the envelope straight away. Results are Sume-hosted signed URLs in data[].url, and usage.cost is the billed USD amount.

What does not exist?

Native streaming is not served: stream: true returns 400 streaming_not_supported. seed and output_compression return 400 unsupported_parameter. For the current field list, see the Image API docs.