Qwen-Image at $0.02 per megapixel vs a flat price per image

fal charges Qwen-Image $0.02 per megapixel, billed rounded up to the next whole megapixel, so a 1024x768 image (0.786 MP) is billed as one megapixel. Sume lists the same family as qwen/qwen-image and bills a catalog price per image, not per megapixel, so the quick way to compare is to read the endpoints record for the model rather than multiply pixels.

fal's price and features are from Qwen-Image on fal and Qwen Image Max on fal, both read 2026-10-03. Sume's billing model is from the Image API docs.

Per megapixel vs per image

On fal, Qwen-Image is described as strong at complex text rendering and precise image editing, with commercial use permitted and LoRA fine-tuning supported. Qwen Image Max is priced at $0.075 per generation on its fal page, which describes it as improving realism and naturalness over earlier versions.

Check the Sume price in one request

Sume publishes one endpoint per model, and the docs say its pricing lines already include Sume's margin. Run the script for the plain model and again for qwen/qwen-image-max, then compare the two lines.

import os
import requests

key = os.environ["SUME_API_KEY"]
url = "https://api.sume.com/v1/images/models/qwen/qwen-image/endpoints"
resp = requests.get(url, headers={"Authorization": f"Bearer {key}"}, timeout=30)
resp.raise_for_status()
body = resp.json()
body = body.get("data", body)
for endpoint in body["endpoints"]:
    for line in endpoint["pricing"]:
        print(endpoint["provider_slug"], line["billable"], line["unit"], line["cost_usd"])

Limits

A flat per-image price means a small render and a large one cost the same on Sume only if the catalog line does not vary by size, and the docs do not say it varies, so verify with usage.cost on a real call. LoRA fine-tuning is a fal feature: Sume's request schema has no LoRA field, and provider.options must be empty, so a custom Qwen LoRA is not something you can run through Sume.