Google Antigravity Quota Explained: Weekly vs 5-Hour Limits, Credits, and What Your Logs Show
Updated November 3, 2026 · first published November 3, 2026
Google's Antigravity documentation says plans not on Google AI Pro or Ultra get a quota refreshed weekly. Pro gets a high quota refreshed every five hours, up to a weekly limit. Ultra gets the highest quota, also refreshed every five hours. Credits cover use beyond the baseline, and Google publishes no quota numbers for any plan.
Published 3 November 2026 · Google pages read 3 November 2026 · local data from one machine, October 2026
What Google says about the plans
From Google's Antigravity pricing page and plans documentation:
| Plan | Quota refresh | What else the page says |
|---|---|---|
| Not on Google AI Pro or Ultra (the Individual plan, listed at $0) | Weekly quota plus a weekly rate limit | Basic weekly rate limits; unlimited tab completions and command requests; Gemini 3.8 Flash as the core agent model |
| Google AI Pro | High quota refreshed every five hours, until the weekly limit is reached | Higher weekly rate limit; purchased or promotional AI credits for use beyond the baseline |
| Google AI Ultra | Highest quota, refreshed every five hours | Highest weekly rate limits; third-party models, which the page does not name |
| Business via Google Cloud | Not stated | Standard and Plus plans from $30 per seat per month; a pay-as-you-go option with consumption pricing |
Limits are described as “primarily determined by available capacity” and as correlating with “the amount of work done by the agent”, so complex tasks use more quota than simple ones. Credits are billed at standard Gemini Enterprise consumption pricing. An AI Credit Overages setting chooses what happens when a model's baseline runs out: Never waits for the refresh, Always spends credits and switches back after the refresh. Exhaustion is tracked per model.
What Google does not say
- Any token, request or credit number for any plan, or the length of the weekly window.
- The size or price of the credit pool.
- Whether Pro includes third-party models, or whether Claude is available, and on which plan.
- Terms for trial or preview models.
Third-party pricing guides fill these gaps with figures such as plan prices and multiples, but they disagree with each other and are not Google's own words. We left them out. For a budget, use Google's pages and the usage screen inside Antigravity.
What our machine recorded
We looked at Antigravity's local data for October. It records activity well and usage badly.
| Question | Local data |
|---|---|
| How active were we? | 235 conversations modified in October, 98 with our own input, 65,349 steps, activity on 28 days |
| Which models ran? | Yes, per conversation, but counts cover the whole life of each conversation, not October only |
| Token counts? | Counter fields exist in an undocumented binary format; we could not verify what each field means, so we publish none |
| Quota balance, credits spent, cost? | Not recorded locally |
| Reset time? | Only as an error message in a command-line log |
The one quota event we found was a command-line error on 3 October at 14:37 (log time): “RESOURCE_EXHAUSTED (429) Individual quota reached… Resets in 2h1m37s”. A reset about two hours away is shorter than a weekly window, which fits the five-hour refresh Google describes for Pro and Ultra. But the logs do not record the plan, so we cannot say which applied.
What this means for a FinOps team
- There is no local ledger. You cannot reconstruct tokens, credits or cost per project from Antigravity's files. Allocation has to come from the product's own usage screen or billing export.
- Quota is not a price. A quota that refreshes every five hours up to a weekly limit is a capacity limit. You only see a bill when you buy credits, at consumption pricing that Google ties to Gemini Enterprise rates.
- Set the overage switch deliberately. Always turns a rate limit into a metered cost; Never turns it into waiting. Pick per person or team.
- Watch trial and preview models. Gemini 4 Argon appears here as a trial model; its API price changes on 1 December (see the price change), and Antigravity's terms for trial models are not stated.
- Compare plans on the same unit. See why provider prices are not comparable.
Limits of this article
- One machine and one user; we do not know which Google plan was active.
- Google's pages may change; this is what they said on 3 November 2026.
- Model-call counts are lifetime counts for conversations touched in October.
Frequently asked questions
Does Antigravity have a free tier?
Google's plans page does not use the word free. It describes users not on Google AI Pro or Ultra as getting a meaningful quota refreshed weekly, plus weekly rate limits, and lists the Individual plan at $0.
How often does the Antigravity quota reset?
Per Google: weekly for plans not on Pro or Ultra; every five hours for Pro (until the weekly limit is reached) and for Ultra. Google gives no exact window length or quota numbers.
What happens when I run out of Antigravity quota?
On Pro and Ultra you can use AI credits beyond the baseline. The AI Credit Overages setting either waits for the refresh (Never) or spends credits and returns to the baseline after the refresh (Always). Exhaustion is tracked per model.
Can I see my Antigravity token usage in local files?
Not reliably. In our October data, conversation files record activity and model names, but no quota balance, credits or cost, and the token counters are in an undocumented binary format.
Related
- Gemini 4 Argon price doubles on December 1
- Why provider prices are not directly comparable
- Claude Code logs overcount tokens 2.5x
- AI API price calendar for Q4 2026
Want this applied to your stack? Bring the provider bills, gateway logs, and top workflows; we will map the cost drivers and savings path. Book a free audit →