What one n=4 image call costs on each Sume image model

On Sume an image call costs the endpoint price times n, because the docs say cost_usd x n is what you pay. A call with the largest n the model allows costs between $0.02 (four Soul images) and $1.055 (four ChatGPT Image 2 images).

Every catalog model except Grok Imagine accepts n up to 4 today. The schema allows up to 10 per request, but per-model ceilings are lower, so read the n range from the catalog before you pin a number.

The table

Prices are per image at the catalog's default settings; the ChatGPT Image rows show the high-quality 1024 rate, and Nano Banana rows the 1K rate. Higher resolution or quality moves the number.

Billing rules that change the math

Generation is all-or-nothing: a completed generation is billed in full, a failed or cancelled one is not billed, and a client that disconnects early is treated as a failed generation. So a four-image call that errors costs nothing, but a four-image call you abandon after a timeout is not a way to get a partial refund of finished work.

Slow configurations (4K, high quality, large n) are the ones most likely to outlast the 30-second blocking window and return 202. Send mode: "async" for those and read the result from the job endpoints.

Compute it from the API

The endpoints record carries the billed price, so a budget check needs no hard-coded table. This script prints the cost of a four-image call for any model id.

import os
import requests

BASE = "https://api.sume.com/v1/images/models"
HEADERS = {"Authorization": f"Bearer {os.environ['SUME_API_KEY']}"}


def call_cost(model: str, n: int) -> float:
    r = requests.get(f"{BASE}/{model}/endpoints", headers=HEADERS, timeout=30)
    r.raise_for_status()
    body = r.json()
    body = body.get("data", body)
    price = body["endpoints"][0]["pricing"][0]["cost_usd"]
    return price * n


if __name__ == "__main__":
    print(call_cost("bytedance-seed/seedream-5-lite", 4))