Scheduled AI runs in Q4 2026: ceiling by cadence at $1 a run

A schedule on Sume carries a generation spend cap for each run. The create-schedule docs set the default at $1.00 per run, let a single run lower it but never raise it above the schedule's cap, and reject 0. Passing null removes the automation ceiling, though wallet, admission and organization limits still apply.

Per-run caps make the quarterly ceiling a count: runs times cap. Cron schedules take a five-field cron expression plus an IANA time zone.

How many runs fit in Q4?

From October 3 to December 31, 2026, there are 90 days inclusive. Mondays between October 5 and December 28 number 13. The table multiplies each cadence by the default cap.

What does the ceiling really mean?

It is the most the automation can spend, not a forecast. A run that makes one 5-second Wan 3.0 clip at 720p costs $0.625, so a weekly schedule of that costs well below $13. The ceiling matters when a prompt loops, a model changes, or an upstream input grows: the cap stops a single run, and the count of runs bounds the total.

A cadence that fires every hour deserves a deliberate cap, not the default. At $1.00 it can still spend $2,160 over the quarter, which is more than a daily schedule with a $20 cap ($1,800).

When is null appropriate?

Rarely. With null, the automation has no per-run ceiling and only the wallet, admission control and organization limits stand in the way. Use it for a run whose cost is bounded by construction, for example a single fixed-price caption job, and keep a numeric cap everywhere else.

To see what $1.00 can actually buy before you choose a cadence, read what a $1 scheduled run can make.