Google Antigravity pricing 2026: tiers, limits, free plan

On May 19, 2026, Google repriced Antigravity’s ladder six months after launch. The top tier dropped from $249.99 to $199.99, a new $99.99 rung appeared in the middle, and separate rate limits for different Gemini models merged into one pool. Google announced all of it on the Antigravity blog, matter-of-factly, as “Changes to Antigravity Plans.”

That event is the honest summary of this product’s economics. The prices are ordinary. What moves is everything around them: rate limits, model access, and quotas that Google’s own documentation reserves the right to change. Budgeting for a tool like that is a different job than reading a price list.

Which is the lens this guide applies. CloudZero is the AI ROI company, so every tier here gets priced the way a finance team would price it: what it costs, what actually governs the bill, and how you’d defend the spend at renewal. Most teams arrive at this decision with the harder problem already on the books: three or four coding agent subscriptions, each approved separately, none mapped to a team or shipped output.

If that’s you, schedule a demo to see agent spend allocated by team and tool first. The right Antigravity tier is obvious once you know who actually uses what.

Google Antigravity pricing: key facts

What is Google Antigravity?

Google Antigravity is Google’s agent-first development platform: an IDE, a terminal CLI, and an agent manager that plan and execute multi-step coding tasks rather than autocompleting lines. It launched in November 2025, absorbed the retired Gemini CLI in June 2026, and expanded into a full agent platform with Antigravity 2.0 at I/O 2026.

The lineage matters for anyone arriving from Google’s older tools. Gemini CLI stopped serving free and individual Google AI accounts on June 18, 2026, and the Antigravity CLI is its replacement, sharing the same agent harness as the IDE. Docs and tutorials that recommend Gemini CLI for individual use now describe a discontinued product.

The Antigravity IDE runs Gemini models as the core agent brain, with third-party models appearing at some tiers. That phrase, at some tiers, is doing careful work, and the free plan section explains why.

Antigravity 2.0, the I/O 2026 release, widened the product from a coding IDE into a platform for building and managing cohorts of agents, which is why its pricing question has grown past developer tooling. A tool that runs one agent is an engineering line item. A platform running cohorts of them is a workload, and workloads get budgeted, forecast, and reviewed like any other consumption spend.

How much does Google Antigravity cost in 2026?

Google Antigravity costs $0 on its Individual tier, and there is no Antigravity-only subscription above that. Paid capacity comes bundled with a Google AI plan: Pro at $19.99 a month, Ultra 5x at $99.99, and Ultra 20x at $199.99. Organizations skip the ladder entirely and pay consumption rates through Google Cloud.

Three structural facts separate this from a normal pricing page. The prices live on Google’s AI subscriptions page, the quota rules live in the Antigravity plans documentation, and the Antigravity pricing page itself lists tiers without most of the dollar figures. No single Google page shows the whole model, which is a large part of why searchers land on Reddit threads instead.

Is Google Antigravity actually free?

Yes, Google Antigravity is genuinely free on the Individual tier, with an asterisk that grows over time. The tier is $0, generally available rather than a preview, and includes unlimited Tab completions and unlimited Command requests. The constraint is the weekly agent quota, which caps how much autonomous agent work you get before the meter stops you.

Two of those allowances are genuinely unlimited, and it is worth knowing which. A Tab completion is the inline code suggestion the editor offers as you type, accepted with the Tab key. A Command request is a natural-language instruction you issue inside the editor to change existing code: rename this, extract that, convert this loop. Neither draws on the agent quota. The weekly cap applies only to autonomous agent runs, where Antigravity plans and executes a multi-step task on its own.

The asterisk: Google has cut the free tier’s limits repeatedly since the November 2025 launch, always downward. A free allowance that covered real daily work at launch covers less of it now. Free remains a legitimate evaluation tier and a light-use tool. Building a team workflow on it means building on terms one blog post can shrink.

And here Google’s own pages disagree with each other. The Antigravity pricing card lists the free tier’s models as Gemini 3.8 Flash, Gemini 3.7 Flash, Gemini 3.6 Flash, Gemini 3.1 Pro, Claude Sonnet and Opus 4.6, and gpt-oss-120b. The plans documentation lists third-party model access as an Ultra benefit. Until Google reconciles them, trust the model selector in your signed-in account over either page, including over this one.

How do Antigravity’s rate limits work?

Google Antigravity meters usage on two clocks. On paid plans, quota refreshes every five hours until a weekly ceiling stops it, per Google’s plans documentation; the free tier refreshes weekly, full stop. Either way, unused allowance does not roll over. More unusual is what the meter counts. The limits track the amount of work the agent performs, not the number of prompts you send, so an identical prompt can consume very different amounts of quota depending on how much planning, execution, and retrying the agent does behind it.

Google is unusually candid about the other variable. The plans documentation states that baseline rate limits are “primarily determined to the degree we have capacity.” Your allowance is partly a function of Google’s available compute in a given period, which is why users report limits tightening without a pricing announcement.

One number is missing from every Google page. There is no published absolute quota figure for any Antigravity tier, so the 5x and 20x multipliers are relative to a Pro baseline Google does not quantify.

For an individual, that’s an annoyance. For a team budgeting the tool, it’s the whole story: the unit you’re buying is elastic on both sides. The May 2026 restructuring proved the point by merging the separate Gemini Flash and Gemini Pro limits into one pool, changing what a subscription bought mid-cycle for everyone.

The practical adjustment is to plan in ranges rather than fixed capacity. Track how far into the week each developer’s quota lasts for a few cycles, treat the worst observed week as the planning baseline, and hold upgrades until the ceiling binds for consecutive weeks rather than once. A quota that ran out during one heavy refactor is signal; a quota that runs out every Thursday is a tier decision.

What are AI credits, and when do overages kick in?

AI credits are Google Antigravity’s overage mechanism. When a Pro or Ultra subscriber exhausts the baseline quota for a model, purchased or promotional credits can cover additional usage, consumed at standard Gemini Enterprise Agent Platform consumption rates rather than at a flat subscription price.

A user setting called AI Credit Overages controls the behavior: never dip into credits, or allow them once the baseline runs dry. That toggle is worth setting deliberately, because it’s the line between a fixed $19.99 line item and a subscription with a usage-based tail. Heavy agent users who leave overages on have converted a flat fee into metered spend without changing anything else.

A free CloudZero cloud cost assessment shows where usage-tailed subscriptions like this already sit across your stack, which is worth knowing before the credits setting gets flipped team-wide.

What do teams pay for Antigravity?

Organizations don’t buy seats on the consumer ladder. Team access runs through the Gemini Enterprise Agent Platform on Google Cloud, priced by consumption under Google Cloud’s billing model, with project-level integration and pooled consumption billing in place of per-user tiers.

Consumption pricing cuts both ways for a finance team. It scales cleanly with real usage and avoids paying for idle seats, the classic failure of per-seat AI subscriptions. It also means the monthly number is a forecast, not a fact, and forecast accuracy depends entirely on knowing which teams drive the agent hours. Nobody quotes a list price; your Google Cloud agreement sets the terms.

A practical way to enter that negotiation: run a measured pilot first. Put one team on paid individual plans for a month, capture how much agent work they actually delegate and what the credit overages look like, then price the organization route against that observed baseline rather than against a vendor estimate. Consumption models reward buyers who arrive knowing their own usage curve, and punish the ones who guess.

How does Antigravity pricing compare to Claude Code and Cursor?

At the entry tier the three cost nearly the same, and the models diverge from there. Antigravity’s paid ladder starts at $19.99 through Google AI Pro. Claude Code comes bundled with Claude Pro at $20 a month, with Max tiers at $100 and $200. Cursor’s Pro plan is $20, with a $60 Pro+ rung in the middle and a $200 Ultra tier at the top.

The sticker prices converge so closely that they’ve stopped being the decision. What differs is forecastability, and there Antigravity is the hardest of the three to plan around: capacity-based limits with a documented history of changing, versus metering a team can model.

Antigravity vs. Claude Code on cost

Comparing Antigravity vs. Claude Code on money alone: entry costs match at about $20, both top out at $200, and Antigravity adds a free tier Claude Code lacks entirely. The structural difference is what the money buys. A Claude Pro subscription sets session-based limits per plan tier, published and stable enough to budget against.

Antigravity’s capacity-linked quota can deliver a different effective allowance from the same $19.99 in different weeks. For a solo developer sampling agents, Antigravity’s free tier wins the trial. For a team standardizing on one agent, Claude Code’s predictability is itself a feature with a dollar value.

Antigravity vs. Cursor on cost

The Antigravity vs. Cursor math is a subscription against an allowance. Cursor’s Pro plan at $20 includes a monthly model-usage allowance with usage-based billing past it, so heavy months show up as overage you can see coming. Antigravity’s equivalent pressure valve is AI credits at consumption rates after the quota empties.

Cursor is an IDE-first product with agent features; Antigravity is agent-first with an IDE attached. Teams already living in Cursor rarely save money switching, while teams betting on autonomous agents get more agent surface per dollar from Antigravity, with the forecasting caveat that runs through this whole guide.

One more coordination note for Google-stack teams: Antigravity access is separate from Gemini’s consumer and API pricing and from Vertex AI, even though all three draw on the same model family. Three Google AI bills can now arrive for one team, each metered differently. The rest of the coding-agent market lives in our Codex pricing guide, the Codex vs. Claude Code comparison, and the broader AI pricing guide.

Does the Antigravity CLI cost extra?

No, the Antigravity CLI is included at every tier, free tier included, and draws from the same quota pool as the IDE and agent manager rather than carrying its own meter. One plan governs all three surfaces, so switching between the terminal and the IDE doesn’t change what you pay or how fast the allowance drains.

The CLI matters most to teams migrating from Gemini CLI, which it replaced in June 2026. The migration is free in subscription terms, but the metering changed with it: Gemini CLI’s individual allowances are gone, and the work-based quota described above now governs terminal sessions too.

Which Antigravity tier should you pick?

Match the tier to how much autonomous agent work you actually delegate, not to how good the tool feels in week one. The free Individual tier covers evaluation and light daily use honestly. The $19.99 Pro rung fits a developer who runs agents on real tasks most days and can tolerate an occasional weekly ceiling.

The $99.99 Ultra 5x tier exists for people who kept hitting that ceiling, and Google added it in May 2026 precisely because the gap between $19.99 and the old $249.99 plan left them stranded. The $199.99 Ultra 20x tier is for agent-heavy daily drivers, and at that spend the honest comparison is no longer tiers but alternatives, including whether consumption billing through the organization route beats a personal plan.

Whatever the tier, decide the AI Credit Overages setting at the same time. The tier caps the subscription; the toggle decides whether the subscription is really the cap.

How should a team budget for agentic coding tools?

Treat agentic tools as AI spend with a usage tail, not as seat licenses. Antigravity makes the case sharper than most: work-based metering means the same developer, same plan, same prompt volume can cost different amounts week to week, and the overage toggle can quietly convert flat fees into consumption billing.

The budgeting failure mode is familiar from every wave of AI spend: each $20 approval looks trivial, the credits and upgrades accumulate, and at renewal nobody can connect the engineering tool line to shipped output. Whether the tools are worth it is answerable, but only with per-team, per-tool visibility in place before the question arrives.

That visibility is CloudZero’s job. CloudZero breaks AI spend down by team, tool, product, and feature, connecting subscriptions, credits, and consumption billing to the work they fund, no tagging required. One team running 50-plus models found more than $1M in savings by seeing which spend produced value and which just produced usage. The same discipline in AI spend management and measuring AI ROI turns a moving target like Antigravity into a number finance can defend.

Schedule a demo to see your coding-agent and AI spend by team, or take the self-guided tour.

Frequently asked questions about Google Antigravity pricing

Author Spotlight

Lyne Carolyne

Lyne Carolyne has several years of experience in AI and cloud economics and brings that understanding into the content she creates. Outside work, she's an avid explorer.