Gemini API Tier 1's $250 monthly cap: 250 Omni clips, then a pause

A Tier 1 Gemini API account has a $250 monthly spend cap, and Google's billing page says that once the cumulative total reaches it, service pauses for every project on the billing account until the next cycle. At Omni Flash's roughly $0.10 per second at 720p, $250 buys 2,500 seconds: 250 ten-second clips, or about 41 minutes of finished footage. Upgrading takes $100 paid plus three days.

The tier ladder in numbers

Google's billing page lists Tier 1 as a linked billing account with a $250 monthly cap, Tier 2 at $100 paid and 3 days old with $2,000, and Tier 3 at $1,000 paid and 30 days old with $20,000 to $100,000 or more. Upgrades "usually reflect within 10 minutes" after a payment or when you meet the criteria. The pricing page gives Omni Flash at about $0.10 per second at 720p.

What the cap does not tell you

The cap is a spend ceiling on the whole billing account, not a per-model quota. Veo, Lyria, text and image calls share it, so a heavy text workload can use the room your video batch needed. The billing page also notes that project-level caps can overshoot for about a 10 minute latency period, so the real stop can land a little past the number you set.

Higher resolutions and 4K raise the per-second rate; Google's pricing page snippet I read gives the 720p figure only, so price other resolutions from the live page before you budget them.

How Sume bounds the same run

Sume has no monthly tier cap in its docs. It works from a prepaid, workspace-scoped USD balance: at submit it reserves the estimate, captures it on success and releases it on failure, per Generation admission. If the reservation cannot be made you get 402 insufficient_credits before provider work starts, and you can read the balance with GET /v1/balance (API reference).

The cost of that safety is the premium: Sume bills Omni at provider list times 1.25 per output second, so 250 ten-second 720p clips are $312.50 there against about $250 at Google list. You are paying for the queue, the job envelope with durable artifact URLs and idempotent retries, not for cheaper pixels.

Decision rule

Under 250 clips a month, Tier 1 on the Gemini API is the cheapest route if you can live with the pause behavior at the cap. Above that, or when several teams share one billing account, a balance-bound workspace with a single 402 failure mode is easier to reason about. Either way, put your own spend counter in front of the batch; the vendor cap is the last line of defense, not the plan.