GPT Image 2.5 API: OpenAI's two endpoints versus Sume's one route

OpenAI serves GPT Image 2.5 on two endpoints, v1/images/generations and v1/images/edits, and lists no others for Flare or Sunburst. Sume exposes one route, POST https://api.sume.com/v1/images: send only a prompt to generate, or add input_references to edit. The model ids change to openai/gpt-image-2.5 and openai/gpt-image-2.5-sunburst.

The differences that matter when porting code are below. Sources are OpenAI's Flare model page and image generation guide plus Sume's Image API docs, read 2026-09-29.

How do the two APIs line up?

What do I change in an existing OpenAI call?

Point at https://api.sume.com/v1/images, use your Sume key as the bearer token, change the model id, and read data[].url instead of decoding image bytes. Fields Sume does not serve return an error: stream: true gives 400 streaming_not_supported, and output_compression gives 400 unsupported_parameter.

curl -X POST "https://api.sume.com/v1/images" \
  -H "Authorization: Bearer $SUME_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "openai/gpt-image-2.5",
    "prompt": "a hand-drawn map of a fictional island",
    "quality": "medium"
  }'

Does Sume bill the way OpenAI does?

The token rates match OpenAI's page ($5 text input, $8 image input, $30 image output per million tokens), but Sume's per-image amounts include its margin and are shown in usage.cost on the response. See GPT Image 2.5 API price for reserve figures.