Antigravity pricing: one number on the page, and four adjectives

Contents

How much does Google Antigravity cost?

Antigravity’s pricing page contains exactly one number. That number is $0.

G

There are four cards on it. The first says For Individuals, $0/month. The other three — Google AI Pro, Google AI Ultra, and an Organization plan via Google Cloud — carry no price at all. You are expected to go somewhere else to find out, and the somewhere else is a Google One subscription page on a different domain.

Read there, in a US browser, the actual ladder is this:

Plan	Price	What Antigravity calls your allowance
For Individuals	$0/month	”Basic weekly rate limits”
Google AI Plus	$4.99/month	”Limited”
Google AI Pro	$19.99/month	”Expanded”
Google AI Ultra 5x	$99.99/month	”Higher”
Google AI Ultra 20x	$199.99/month	”Highest”
Organization	Via Google Cloud	Consumption-based

Google’s own announcement rounds those to $20, $100 and $200. Both are accurate; the pennies come from the Google One page and the round numbers from Antigravity’s blog.

The part Google will not tell you

Here is the finding, and it is verifiable in about thirty seconds on Google’s own comparison table.

That table quantifies almost everything it sells. Google Flow Credits: 200, 1,000, 10,000 and 25,000 a month across the four tiers. Google Flow Music Credits: 3,000, 10,000, 30,000 and 30,000. Google Cloud credits: $10, $40 and $100 monthly. Exact figures, per tier, no ambiguity.

Then you reach the Antigravity section. It has one row, labelled Agent requests, and its four cells read:

Limited · Expanded · Higher · Highest

Four adjectives. No number anywhere. On the same page, in the same table, where music credits are priced to the individual credit.

Antigravity’s own documentation does the same thing. It describes the quota as “the highest, most generous quota” on Ultra, “high, generous quota” on Pro, and “meaningful quota” on the free tier. The only numeric detail is the refresh interval: paid quotas reset every five hours, the free tier’s resets weekly.

One trap if you go and check this yourself. The row immediately above Antigravity’s is Flow Music Credits, which ends in “30,000/mo”. Read the page as flat text — which is what a scraper or a language model does — and that 30,000 looks like it belongs to Antigravity. It does not. Every figure on this page was extracted from the table’s row structure rather than its reading order, specifically because that adjacency is so easy to get wrong.

The only real number: 5x and 20x

There is exactly one quantified statement about Antigravity’s limits, and it is in Google’s plan-change announcement rather than on any pricing page.

The $100/mo Ultra plan will have a rate limit of 5X worth of tokens compared to the $20/mo Pro plan and the $200/mo Ultra plan will have a rate limit of 20X worth of tokens compared to the $20/mo Pro plan.

That is genuinely useful and also incomplete in a specific way. It tells you the tiers relative to each other. It does not tell you what the $20 tier is, so a multiplier of an unknown is still an unknown. If Pro’s quota suits you, Ultra gives you five times as much of a thing you can measure. If it does not, 5x of “not enough” is not a number you can plan against.

What changed in May 2026

Most of the confusion on this topic dates from one restructure, effective 19 May 2026:

	Before	After
Entry paid tier	$20 Pro	$20 Pro, unchanged
Middle tier	did not exist	$100 Ultra, 5x Pro’s tokens
Top tier	$250	$200, 20x Pro’s tokens
AI credits	Included in base plans	Removed — overage only

The price cut is the part that got written about. The credits change is the part that matters more: credits stopped being something your subscription includes and became the mechanism that charges you when your quota runs out. Google increased Gemini quotas in exchange, so it is not a straightforward downgrade — but it does mean anything written before May 2026 describes a different product.

That restructure is why published guides currently offer $200, $249.99 and $19.99 as competing answers, along with at least one describing the whole thing as a free public preview. They are not all wrong. They are quoting different moments and mostly not saying which.

What the free tier actually gives you

This is the tier most people should start on, and it is more generous than the marketing implies.

Free tier includes
Price	$0/month, no card
Agent models	Gemini 3.5 Flash, Gemini 3.1 Pro, Gemini 3 Flash, gpt-oss-120b
Third-party models	Claude Sonnet and Opus 4.6
Tab completions	Unlimited
Command requests	Unlimited
Product features	All of them, including Scheduled Tasks and the CLI
The catch	”Basic weekly rate limits”, quantity unstated

Two things stand out. Claude models on a free Google tier is not what you would predict from either company, and it is on the page. And all product features are included rather than gated — the paid tiers buy throughput, not capability.

I can put one measured data point against it. I gave Antigravity a six-requirement brief and it built a complete working app in about two and a half minutes, on this tier, for nothing — and it wrote and ran its own unit tests while it was at it. The free tier is not decorative. Details are in the Antigravity review.

What I cannot tell you is how many of those you get in a week, because nobody publishes it.

What the rate limits actually mean

Since the quotas are unpublished, the mechanics are the only thing you can plan around. Here is what Antigravity’s documentation does state.

	Free	Pro and Ultra
Quota described as	”Meaningful quota"	"High, generous” / “Highest, most generous”
Refresh interval	Weekly	Every five hours
Weekly ceiling	Yes	Yes, higher
Tab completions	Unlimited, not metered	Unlimited, not metered
When you run out	Wait for the reset	Credits, if overages are on

The refresh interval is the most practically useful fact on this page, and it is easy to miss.

A five-hour refresh changes how you work. On a paid tier, running out at eleven in the morning means you are back before the end of the day. Exhausting a weekly allowance on the free tier on a Monday means Monday was your week. That difference matters more to daily rhythm than the size of the quota does, and it is the clearest argument for the cheapest paid tier over the free one.

Tab completions are free of the meter on every tier. If most of your usage is autocomplete rather than agent runs, your consumption looks nothing like someone driving agents all day, and the free tier will last far longer than any warning suggests.

There is a weekly ceiling above the five-hourly one. So a paid plan is not simply “five-hourly refills forever” — you can still hit a wall that lasts until the week turns. Google does not say where either boundary sits.

Antigravity’s price against the alternatives

The comparison people actually run is against the tools they are already paying for, so here it is with the figures I have verified for each.

Tool	Free tier	Entry paid	Metered usage on top?	Allowance published?
Antigravity	Free indefinitely	$4.99 Plus / $19.99 Pro	Credits, overage only	No
Cursor	Hobby, no card	$20 Pro	Yes, in arrears	Yes, $20–$400 by tier
Claude Code	None	$20 Claude Pro	Or metered API	Partly
Codex	Bundled with ChatGPT	Included	No	Partly
Lovable	5 credits a day	$25 Pro	No, credits expire	Yes, per-action rate card
Replit	Daily Agent credits	$20 Core	Checkpoints by effort	No

Two observations that a straight price row hides.

At the entry tier, four of these cost about the same. Antigravity Pro, Cursor Pro, Claude Pro and Replit Core all sit at $20 give or take a penny. Price is not the differentiator at this level; what you can find out about what you are buying is.

Antigravity and Replit share the same flaw from opposite directions. Replit shows you a percentage of an unstated total running down. Antigravity does not show you a meter at all. Cursor tells you precisely what everything costs and then bills you a month late — which is how I ended up paying $510.70 across eleven months on a plan advertised at $20. Lovable is the only tool in this set that publishes what an action costs before you take it.

So the fair summary is that Antigravity is the cheapest genuine entry point in the category and the least forecastable. Whether that trade works depends entirely on whether your usage is near a boundary you cannot see.

The Organization plan

Worth a short note because it is new and mostly undocumented publicly.

Antigravity now lists an Organization plan via Google Cloud, marked as rolling out to select customers. What the page states is that access falls under Google Cloud Terms of Service, integrates with a Google Cloud project, and uses consumption-based API pricing through the Gemini Enterprise Agent Platform rather than a seat subscription. It is also included in some Gemini Enterprise app subscriptions.

That is a genuinely different commercial model from everything above: metered against a cloud account rather than a flat monthly fee, which for a team is usually more predictable rather than less, because it lands in billing you already monitor.

No pricing is published for it, and I have not used it, so treat this as a pointer rather than a recommendation. If you are evaluating for a team already on Google Cloud, that route is the one to ask your account rep about rather than buying individual AI Ultra seats.

Should you pay for it?

The honest sequence is: start free, and let the ceiling tell you.

Stay on free if you are evaluating, building occasionally, or working on something small. It costs nothing, includes every feature, and completed a real app for me. There is no trial clock to beat.

Your situation	Tier
Evaluating, or building occasionally	Free, costs nothing, every feature included
Free is nearly enough	Plus, $4.99, the cheap experiment before the $20 one
Daily driver	Pro, $19.99, what Google’s own multipliers measure against
You hit Pro’s ceiling routinely	Ultra, $99.99 (5x) or $199.99 (20x)
Buying for a team on Google Cloud	Organization, consumption-based, ask your rep

Google AI Plus at $4.99 is the cheapest way to move up one step, and it is easy to miss because Antigravity’s own materials talk about Pro and Ultra. Plus does include Antigravity — the comparison table gives it “Limited” rather than a blank — so if the free tier is nearly enough, this is a five-dollar experiment before a twenty-dollar one.

AI Pro at $19.99 is the sensible daily-driver tier and the one Google’s own multipliers are measured against. It is also priced identically to Cursor Pro and Claude Pro, which makes the comparison easy on price and impossible on what you get, since the other two publish their allowances and this one does not.

AI Ultra at $99.99 or $199.99 buys 5x or 20x Pro’s tokens. Worth it only once you have hit Pro’s ceiling often enough to know that multiplying it solves your problem.

On a student discount, I have not verified one for Antigravity specifically. Google has historically offered discounted or free AI Pro access to verified students in some countries, and since Antigravity rides on those same Google One plans, any such offer would carry it. I am not going to quote terms I have not read, so check the Google One page from your own account before assuming it applies.

Turn off AI Credit Overages unless you want the bill to be open-ended. Since May 2026 credits exist to charge you past your quota, and the setting controlling that has a “Never” option. It is the only spend control in the product.

Two things worth knowing before you commit

Google has re-tiered a free developer product here before, and recently. Gemini CLI was retired on 18 June 2026 and its users moved to Antigravity CLI, with the free allowance going from roughly a thousand requests a day to roughly twenty. That is not a prediction about Antigravity’s free tier. It is the most relevant precedent available, from the same company, in the same product family, in the same year.

The price you see depends on where you are. Google One’s plan page is served by IP address, and ?hl=en&gl=US does not override it — a request from outside the US returns local currency. That is why a Reddit thread on this topic quotes “$29/month” for Pro: it is Kenyan shillings converted after bank fees. Every dollar figure on this page came from the /intl/en_us/ path specifically to avoid that, and if you are comparing your own bill against this article, check which currency your account is being quoted in first.

The final word

Antigravity’s pricing is genuinely good and its pricing disclosure is the worst of any tool I have looked at this year.

The good part is real: a free tier that runs indefinitely, includes every feature, ships Claude models on a Google plan, and built a working application for me without a credit card. If you want to find out whether agent-first development suits you, this is the cheapest way to do it that exists.

The bad part is that you cannot plan with it. A pricing page with one number on it, two tiers described in identical words, a comparison table that prices music credits to the unit and your agent requests in adjectives, and documentation whose most precise statement is “meaningful”. The only quantified fact anywhere is a multiplier of an unpublished baseline.

So the recommendation is shaped by the disclosure rather than the price. Start on free, because it costs nothing and it works. Move to Plus at $4.99 or Pro at $19.99 when you hit the wall, since hitting the wall is the only way you will ever learn where it is. And set the overage control before you need it, because the one thing Google is precise about is how to charge you when the quota you cannot see runs out.