Sume catalog price bands: estimated, minimum and maximum USD cents

What do estimated, minimum and maximum mean in the Sume catalog?

Each entry in model_pricing from GET /v1/catalog carries estimated_usd_cents, minimum_usd_cents and maximum_usd_cents. The estimate is the price of a default request, the minimum is the cheapest request the endpoint accepts, and the maximum is the most one call can reserve.

All three are whole cents in USD with Sume's 1.25 house margin already applied, and billing_unit is always usd_cent. They are not three prices you pick between. They bracket what one request can cost, and the real charge for your request lands somewhere inside the band.

How do I list the bands for every model?

The call needs only an API key. This script prints one line per priced model, with minimum, estimate and maximum in cents.

import json, os, urllib.request
req = urllib.request.Request("https://api.sume.com/v1/catalog",
    headers={"Authorization": "Bearer " + os.environ["SUME_API_KEY"]})
data = json.load(urllib.request.urlopen(req))["data"]
for item in data:
    for p in item.get("model_pricing", []):
        print(p["model"], p["minimum_usd_cents"],
              p["estimated_usd_cents"], p["maximum_usd_cents"])

Which band should I budget with?

Use the estimate for a rough per-job figure, the maximum for a wallet check, and the usage ledger for what you actually spent. The reservation stored at submit time is the estimated billable amount for your specific request, rounded up to whole cents. It is captured when the job completes, and failed or pre-generation-canceled jobs are refunded, as described in generation admission.

The catalog band is an endpoint-wide range, so a default estimate can be far from your request. Video 1.0 shows 232 cents as its estimate but a 1,742 cent maximum, and the Video Router shows 50 cents against 4,286.

What do the bands look like for common models?

These values come from the Sume catalog as of this read.

What do the bands not tell me?

A band does not give the cost of your request. For per-second endpoints you still multiply the rate by your length, and the rate card lists each rate. The band also does not set a cap for you: Sume does not turn the maximum into a spending limit, it only shows how much a single call can hold.

For the billing view after the fact, read the job usage summary and quote its debited figure rather than adding rows yourself. If you want a calculator-style walkthrough first, start with the pricing SKU guide.

Can I turn the bands into a pre-flight check?

Yes, with two comparisons. First, read GET /v1/balance and compare it to the maximum of the endpoint you are about to call, which tells you whether any single call could be refused for lack of funds. Second, compare it to your own computed price for the request.

A team running a nightly batch can fetch the catalog once at the start, store the three figures per model, and log them next to the debited amount from the usage summary afterwards. That gives you a record of how far your real spend sat from the estimate, using only numbers Sume returns.

Do not hard-code the cents into your app for longer than a session. Prices and margins can change, and the catalog is the place Sume publishes the current ones.