GPT Image 2.5 output_compression returns 400 on Sume: use Pillow

Sume does not serve output_compression: a request that sends it returns 400 unsupported_parameter. OpenAI's guide documents a 0 to 100 compression setting for JPEG and WebP output, but on Sume you ask for output_format of png, jpeg or webp and shrink the file yourself afterwards, for example with Pillow.

What each side does

Both rows are from the pages I read.

Why a 400 and not a silent drop

The Sume docs call output_compression and seed schema-only: no model advertises them, so the catalog rejects them instead of dropping them. That is deliberate. It means a silent quality change never reaches you, but it also means a client ported from OpenAI's samples fails until you remove the field.

A Pillow recipe

This script requests a PNG, downloads the Sume-hosted URL and saves a JPEG at quality 82. It needs pip install pillow requests and a SUME_API_KEY. The quality value 82 is an arbitrary starting point, not a Sume or OpenAI recommendation; compare file sizes at your own settings. The script stops with a message if Sume returns a 202 job.

import io, os, requests
from PIL import Image

H = {"Authorization": "Bearer " + os.environ["SUME_API_KEY"]}
r = requests.post("https://api.sume.com/v1/images", headers=H, timeout=60, json={
    "model": "openai/gpt-image-2.5",
    "prompt": "a product photo of a red kettle on a white table",
    "output_format": "png",
})
r.raise_for_status()
if r.status_code == 202:
    raise SystemExit("slow job (202): read it from GET /v1/jobs/{id}/result")
url = r.json()["data"][0]["url"]
img = Image.open(io.BytesIO(requests.get(url, timeout=60).content)).convert("RGB")
img.save("out.jpg", "JPEG", quality=82, optimize=True)
print("saved out.jpg", os.path.getsize("out.jpg"), "bytes")

One caution

If you need transparency, keep PNG or WebP. JPEG has no alpha channel, so the convert("RGB") step above would flatten it.