GPT Image output_compression: OpenAI has it, Sume returns 400

On OpenAI's API, output_compression takes 0-100 for jpeg and webp output. On Sume it is in the schema but "not served in v1": sending it returns 400 unsupported_parameter. Choose output_format (png, jpeg, webp or svg) and compress the file on your side if you need a smaller size.

OpenAI's behavior is from its image generation guide; Sume's from the Image API docs and the generation code, read 2026-09-30.

What does OpenAI say about output_compression?

The guide says that with jpeg or webp you can specify output_compression to control the compression level (0-100%), and gives output_compression=50 as the example.

What does Sume do with it?

The docs list output_compression and seed as schema fields that no model advertises yet. The code rejects them with the message "No image model advertises this parameter in v1."

How do I still get a smaller image?

Request output_format: "webp" or "jpeg", then recompress after download with your own tooling. The format options are covered in PNG, JPEG or WebP output.

Will this change?

The wording is "in v1" and "no model advertises them yet", so the answer lives in the model's supported_parameters from GET /v1/images/models. If a model starts listing output_compression, that descriptor is the signal; do not rely on this post for later dates.