gpt-image-2.5 quality auto reserves max: holds from $0.22 to $0.89

Sending quality: "auto" with no size reserves $0.8895 per gpt-image-2.5 image on Sume, compared with $0.2224 when you omit quality and $0.0659 when you also pin 1024x1024. The Image API states that auto quality reserves max, and that auto size reserves the upper bound of output tokens.

The reserved amount is a hold at submit, not necessarily the final charge, but a hold you cannot cover returns 402 insufficient_credits before work starts.

Hold by setting

The safe request

Pin both. A low balance then fails because of a number you chose.

import os, requests

r = requests.post(
    "https://api.sume.com/v1/images",
    headers={"Authorization": f"Bearer {os.environ['SUME_API_KEY']}"},
    json={
        "model": "openai/gpt-image-2.5",
        "prompt": "Flat lay of three ceramic bowls",
        "quality": "medium",
        "image_size": {"width": 1024, "height": 1024},
    },
    timeout=60,
)
print(r.status_code)
body = r.json()
if r.status_code == 200:
    for item in body["data"]:
        print(item["url"])
else:
    print(body)

Why it matters for batches

A loop of 1,000 requests that leaves both on auto needs $889.50 of headroom per thousand, against $16.50 when pinned to medium. Concurrent requests each hold their own reservation, so unpinned batches can hit 402 far earlier than the real spend suggests.

Rule of thumb

• Always set quality.
• Always set image_size or a named size you have priced.
• Treat the table above as the hold your balance must cover per in-flight request.
• Read GET /v1/images/models/{model_id}/endpoints for the live pricing line before a large run: prices are catalog data and can change.