Disclosure: some links on this page may be affiliate links; we may earn a commission at no extra cost to you. Read the disclosure.

By the toolist.ai editors. Researched from Google Antigravity's own pages and hands-on tests published by others, not hands-on tested by toolist.ai. Updated October 15, 2026.

Google Antigravity review: pricing, features and what testers found

Google Antigravity is Google's agentic development platform: a desktop app for running several Gemini coding agents at once, a terminal CLI, its own editor plus extensions for VS Code, Visual Studio, JetBrains IDEs, Zed and Xcode, and a Python SDK. It is free with a weekly quota; Google AI Pro at $19.99 a month is the cheapest plan that raises the limits and adds Gemini 4 Argon, and Google AI Ultra costs $99.99 or $199.99 a month.

Our take

Google Antigravity suits developers who want Gemini coding agents in a desktop command center, the terminal or their own editor and can live with quotas Google does not put numbers on: it is free to start, Google AI Pro at $19.99 a month (or $199.99 a year) is the cheapest plan that raises its limits and adds Claude Sonnet and Opus 5.5 and Gemini 4 Argon (rolling out to Pro since October 15, 2026, with its own weekly quota), and Google AI Ultra at $99.99 or $199.99 a month gives 5 or 20 times Pro's quota. In The New Stack's June 2026 test, Jessica Wachtel found the Antigravity CLI could edit files and run tests, mypy and git in scripted mode where the Gemini CLI it replaced could not, though it was slightly slower on a trivial prompt and signs in through a browser. After a week with the Antigravity 2.0 app in June, running the older Gemini 3.5 model, XDA's Parth Shah found it moved confidently through multi-file work but still needed follow-ups on design details, and put Claude Code ahead only on some reasoning-heavy tasks; testing the original editor in January 2026, Visual Studio Magazine's David Ramel got a plainer site than Cursor's and an agent that implemented changes it was only asked to suggest. The main catch is usage: Google publishes no quota figures (the free tier refreshes weekly, Pro every five hours up to a weekly limit), users have reported sudden cuts, extra usage needs AI credits bought through Google One, whose price per credit Google's plan and help pages do not state, and on individual plans Google uses your interactions to improve its products, with human review, unless you turn telemetry off.

Based on Google Antigravity's published plans and the hands-on tests linked below, not on a test of our own.

Good for

  • Developers who want to run and review several coding agents at once from a desktop app, the terminal or the editor they already use
  • Google AI Pro or Ultra subscribers who want coding agents inside a plan they already pay for
  • Front-end and full-stack work where the agent checks its result in a browser and leaves plans and recordings to review
  • Former Gemini CLI users who signed in with a personal Google account, free or on Google AI Pro or Ultra, which the Gemini CLI stopped serving on June 18, 2026
  • Trying an agent-first coding tool at no cost before paying

Watch out for

  • Google publishes no quota figures: the free tier gets a quota it calls 'meaningful' that refreshes weekly, Pro a 'high, generous' quota refreshed every five hours until a weekly limit, and Ultra 5 or 20 times Pro's. Usage depends on how much work the agent does, and Google says limits may change to manage capacity; in March 2026 DevClass reported Pro users hitting week-long waits after Google changed its plans.
  • Past the quota, only Pro and Ultra can keep working, with AI credits bought through Google One in packs such as 2,500, 5,000 or 20,000 and charged at standard API prices for the model. Google's public pages give no dollar price per credit, credits may expire, and they cannot be bought on Google AI Plus, in Japan, or by subscribers billed through Apple, Pixel Pass or another third party; an overage setting (Never or Always) decides whether they are spent automatically.
  • On individual plans, free or paid, Google records your use of Antigravity and uses it to improve its products and models, and its employees and contractors may review it, unless you turn off Enable Telemetry in settings; you can email Google to have it deleted. Gemini Enterprise and the paid Gemini API are not used to train Google's models.
  • Google's pages disagree on which plans get which models. Gemini 4 Argon reached Antigravity on October 13, 2026 for Ultra only, as Google's launch post still says; on October 15 Google's Antigravity account said it was rolling Argon out to paid Pro subscribers with a separate weekly-only Argon quota. Antigravity's model table and FAQ list Argon and Claude Sonnet and Opus 5.5 for paid Pro (not trials) and Ultra, while Google One's plan comparison still shows Argon in Antigravity for Ultra only and Antigravity's plans page lists third-party models as an Ultra benefit. Check the model picker before upgrading for a particular model; Claude Sonnet 4.6, Claude Opus 4.6 and GPT-OSS 120B are removed from Antigravity on November 2, 2026.
  • Google AI Plus (from $4.99 a month) gets the free tier's access in Antigravity's plans and model pages (one overview page says otherwise) and cannot buy AI credits, so Pro is the first plan that raises Antigravity's limits. You can cancel Google AI plans anytime, but there are no refunds for partial billing periods unless the law requires one.
  • Individual plans need a personal Google Account in a supported country and age 18+ (Workspace accounts may fail to sign in). Using your Antigravity sign-in from third-party tools such as OpenClaw, Claude Code or OpenCode breaks the terms and can get the account suspended, and Google says wrapping agy's headless mode in other agents or automating account rotation is not supported on consumer plans.
  • The Antigravity CLI is closed-source and replaced the open-source Gemini CLI for free, Pro and Ultra users on June 18, 2026 without full feature parity; it signs in through a browser, which The New Stack's tester found awkward for headless servers. Agents can run terminal commands and drive a browser, so Google advises keeping approvals or the sandbox on outside disposable environments.

What Google Antigravity does

Price

Free plan: Free with a personal Google Account (18+): every Antigravity surface and feature, unlimited Tab completions and Command requests, Gemini 3.8 Flash plus Claude Sonnet 4.6, Claude Opus 4.6 and GPT-OSS 120B until November 2, 2026, and a weekly quota Google does not state as a number; no Gemini 4 Argon and no AI credits for overage, and Google uses your interactions to improve its products unless you turn telemetry off.

PlanBilled monthlyBilled yearly, per monthIncludes
Free free not listed Personal Google Account, 18+ (Google AI Plus subscribers get the same in Antigravity): the app, CLI, IDE, IDE extensions and all product features; unlimited Tab completions and Command requests; Gemini 3.8 Flash, plus Claude Sonnet 4.6, Claude Opus 4.6 and GPT-OSS 120B until November 2, 2026; a quota refreshed weekly with no stated figure; no AI credits for overage
Google AI Pro $19.99 $16.67 Higher Antigravity limits: a quota refreshed every five hours until a weekly limit (no figures); on paid subscriptions, not trials, Claude Sonnet 5.5, Claude Opus 5.5 and Gemini 4 Argon (rolling out since October 15, 2026, with a separate weekly-only Argon quota); Claude Sonnet 4.6, Claude Opus 4.6 and GPT-OSS 120B until November 2, 2026; AI credits for overage, bought through Google One; plus the Gemini app, 5 TB of storage and Google's other AI Pro benefits; $199.99 billed yearly
Google AI Ultra 5x $99.99 not listed 5 times Google AI Pro's usage quota in Gemini and Antigravity, prioritized traffic and first access to new models; Gemini 4 Argon, Claude Sonnet 5.5 and Claude Opus 5.5; AI credits for overage; storage from 20 TB and Google's other AI Ultra benefits
Google AI Ultra 20x $199.99 not listed 20 times Google AI Pro's usage quota in Gemini and Antigravity (Google's highest Antigravity limits); otherwise as Ultra 5x
Gemini Enterprise Standard / Plus not listed not listed Through Google Cloud, contact sales: Google lists Standard and Plus from $30 a seat a month and does not say on these pages whether that needs a yearly commitment; the Antigravity app, CLI and IDE extensions (not the standalone IDE) under Google Cloud terms; Gemini models draw on pooled seat credits ($10 a month a Standard seat, $15 a Plus seat), and Claude Sonnet and Opus 5.5 are billed per token once an admin enables them; SSO, VPC Service Controls and data residency; code and prompts not used to train Google's models
Gemini Enterprise pay-as-you-go not listed not listed No seat fee: a Google Cloud project with an invoiced billing account, billed by consumption, with spend caps across API and subscription plans
Gemini API (Antigravity SDK) not listed not listed Pay per million tokens, input/output: Gemini 3.8 Flash $0.75/$3.75 through December 31, 2026 ($1.50/$7.50 from January 1, 2027), with a free tier whose content Google uses to improve its products; Gemini 4 Argon $2/$10 through November 30, 2026 ($4/$20 from December 1), paid tier only

Prices last checked: 2026-10-16 on Google Antigravity's pricing page. Full breakdown: Google Antigravity pricing.

Price history

No price change since we first recorded these prices on October 15, 2026. We re-check them every week and list each change here.

What hands-on testers found

We have not tested Google Antigravity ourselves. These are the published tests we found that used it hands-on, newest first; follow the links for the full reviews.

XDA Developers · Parth Shah · June 30, 2026

What they tested: A week of agentic coding in the Antigravity 2.0 app with Gemini 3.5, a model Antigravity no longer lists: turning a draft SaaS landing page into a product site from one detailed prompt. He also draws on his June 15 XDA comparison, in which he built the same personal knowledge-management web tool with Antigravity (Gemini 3.5), Claude Code and Codex.

  • One detailed prompt was enough for the agent to work through the site's files by itself, reworking the layout, adding the form, making the page behave better on phones and writing up what it changed; the tool came to feel to him like somewhere to hand out and check work rather than type code.
  • It still got details wrong, and he had to follow up on spacing, hierarchy and overall feel himself.
  • On the knowledge tool, Gemini 3.5 in Antigravity delivered a polished interface with tags, tasks, a knowledge graph and notes; he still rates Claude Code better on some complex, reasoning-heavy work, but no longer by much.

The New Stack · Jessica Wachtel · June 20, 2026

What they tested: Antigravity CLI 1.0.9 against Gemini CLI 0.47.0 on the day Google switched personal accounts over: timed trivial prompts, a scripted type-hinting task on a Python repo, migrating a Gemini CLI extension, and piping code to the CLI from a script.

  • Run without prompts in scripted mode, the Antigravity CLI finished the Python task in about 2 minutes 11 seconds: it set up a virtual environment, ran pytest, mypy and git, and correctly reported the file was already fully typed, where the Gemini CLI could not write files or run shell commands in that mode.
  • Google's 'snappier' claim did not hold on a trivial prompt (a median of about 3.97 seconds against 3.2 for the Gemini CLI), but its timings were more consistent, it hit none of the 503 errors that used up her free Gemini quota, and it explained each step in plain language.
  • A Gemini CLI extension migrated and ran as a skill once she packaged it properly, but the browser sign-in on first run is an obstacle on headless servers; she judged Antigravity the better tool while noting it was a first pass on switch-over day.

Visual Studio Magazine · David Ramel · January 26, 2026

What they tested: The original Antigravity editor, about two months after its November 2025 debut in preview and before the Antigravity 2.0 app, the CLI and the extensions, against Cursor and Windsurf: building and then redesigning the same four-page static website in plain HTML and CSS.

  • Its site was plainer than Cursor's, the most polished of the three, with fewer visual flourishes, though it added light and dark mode, mobile navigation and scroll animations and put correctness and accessibility first.
  • It wrote planning artifacts (an implementation plan, a task list and a walkthrough) that he could comment on line by line to steer the agent, which neither Cursor nor Windsurf offered.
  • Asked only to suggest improvements, it went ahead and implemented them, and previewing the site meant installing a Chrome extension where the other editors simply opened the file.

What users say

Test result

toolist.ai has not run its own test of Google Antigravity; the findings above come from the published tests linked. See how we research and test.

Best coding tools · Google Antigravity pricing

Do you make Google Antigravity? Claim this listing to correct it; it's free. To see whether ChatGPT, Claude, Gemini and Perplexity recommend it, and how many readers we send you: toolist.ai for AI companies.

Sources