Claude
Anthropic's AI assistant built for long-context analysis, coding, and safe reasoning
Google's agent-first development platform — desktop app, IDE, CLI, and SDK sharing one agent harness across Gemini, Claude, and GPT-OSS models
Google Antigravity is Google's agent-first development platform, launched in public preview in November 2025 alongside Gemini 3 and expanded at I/O 2026 into four surfaces: a standalone desktop app, the original VS Code-based IDE, a Go CLI that replaced Gemini CLI on 18 June 2026, and an SDK. It stays free for individuals at $0 under a metered preview quota; higher limits ride on Google AI subscriptions — Pro at $19.99/month, Ultra from $99.99/month at roughly 5x Pro quota, and a $200/month top tier at roughly 20x. It runs Gemini 3 models plus Anthropic Claude and GPT-OSS-120B. For raw coding quality Claude remains the stronger pick; Antigravity competes on price and parallel agent orchestration.
Google Antigravity is Google’s agent-first development platform, announced in November 2025 alongside Gemini 3 and free for individuals in public preview ever since. The original product was an IDE — a modified VS Code fork built by the team Google brought over from Windsurf — whose defining feature was the agent manager: a dedicated surface where you spawn, orchestrate, and observe multiple agents working asynchronously across different workspaces, rather than chatting with one assistant in a sidebar. Agents work across the editor, terminal, and a controlled Chrome instance, so an agent can change a UI and then open the browser to verify the change actually rendered. It has never been Gemini-only: Claude Sonnet and Opus and the open-weights GPT-OSS-120B sit in the model picker next to Gemini 3 Pro and Flash, which means the reasoning-heavy agents in a run can be pointed at Claude while cheaper mechanical work goes to Flash.
At Google I/O 2026 the product widened from an IDE into four surfaces sharing a single agent harness: Antigravity 2.0, a standalone desktop app built around an Agent Teams panel where subagent squads run in parallel, each in its own sandbox; the original Antigravity IDE, still shipping; the Antigravity CLI, a Go rewrite for keyboard-first and remote work; and the Antigravity SDK for hosting custom agents on your own infrastructure. That consolidation had a casualty — Google retired its own Gemini CLI, which stopped serving free, AI Pro, and AI Ultra requests on 18 June 2026, pushing those developers onto the Antigravity CLI while enterprise Gemini Code Assist licences carried on unchanged. Pricing has moved in the same period. Antigravity still costs nothing for individuals, but the free quota has been tightened repeatedly since launch, higher limits now ride on Google AI subscriptions (Pro at $19.99/month, Ultra from $99.99/month at roughly 5x Pro quota, and a $200/month top tier at roughly 20x), and AI credits arrived in March 2026 as a pay-as-you-go overage currency at $25 per 2,500 credits.
Antigravity fits developers who want serious multi-agent orchestration without a per-seat editor bill, and who can tolerate preview-stage instability in exchange. The zero-cost individual tier makes evaluation genuinely free, and the model picker means you are not locked to Google’s own models. The catch is commitment: this is still a public preview with no standalone price, quotas that have been cut without much warning, and a vendor that demonstrably retires developer tools — so weigh it accordingly before a team standardises on it.
Starting price: $0 · Free tier: yes · Model: freemium
Price history tracked from June 2026
| Plan | Price | Includes |
|---|---|---|
| Individual (Public Preview) | Free | No cost for individuals during the public preview · Metered by usage quota, not a dollar cap · Quota refreshes on a rolling ~5-hour cycle up to a weekly ceiling · Free limits have been cut repeatedly since the November 2025 launch · No credit card required |
| Google AI Pro | $19.99/mo | Baseline paid Antigravity quota · Bundled into the Google AI Pro subscription, not sold standalone · Includes 5 TB storage and wider Gemini app limits · Eligible to buy pay-as-you-go top-up AI credits |
| Google AI Ultra | From $99.99/mo | About 5x the Antigravity usage limit of the Pro plan · Introduced at Google I/O 2026 as a mid Ultra tier · Top Ultra tier reaches $200/mo at about 20x Pro limits · Top tier was repriced down from $249.99/mo · Priority access to the newest models |
| AI Credits (top-up) | $25 per 2,500 credits | Overage mechanism for AI Pro and Ultra subscribers · Introduced March 2026 as Antigravity's usage currency · Credit-to-token conversion rate is not published by Google |
| Pros | Cons |
|---|---|
| Free for individuals at $0 with no credit card — the cheapest way to evaluate a serious agentic IDE | Still a public preview with no standalone price commitment — free access can be repriced or withdrawn, and the free quota has already been cut hard (roughly 250 requests/day at launch down to about 20/day by December 2025) |
| Not Gemini-only: Claude Sonnet and Opus and GPT-OSS-120B are first-class in the model picker, so you can route the hard reasoning to Claude | Quota terms are vague and have moved without warning; in March 2026 users reported weekly ceilings dropping from hundreds of millions of input tokens to under 9 million, and Google does not publish the credit-to-token conversion rate |
| Paid quota folds into a Google AI Pro or Ultra subscription you may already hold rather than a separate per-seat bill | Google-experiment mortality risk is real and demonstrated — Google killed its own Gemini CLI on 18 June 2026, roughly seven months after Antigravity launched; the same platform risk applies to Antigravity itself |
| Agent manager and Agent Teams handle parallel multi-agent work better than a single-threaded chat pane | On pure coding quality Claude still leads; Antigravity's edge is orchestration and price, not the strongest code generation available |
| Runs natively on macOS, Windows, and Linux, with a Go CLI for keyboard-first and remote work | No bring-your-own-key, local models, or custom endpoints — the model list is curated through Vertex Model Garden |
| Paid quota requires a consumer Google AI subscription tied to a personal Google account, and there is no web-based IDE |
Anthropic's AI assistant built for long-context analysis, coding, and safe reasoning
AI-native code editor that runs autonomous agents to build, test, and review code
Agentic AI IDE with Cascade that writes, edits and runs code autonomously inside your editor
AI coding assistant with inline completions, chat, and autonomous agent mode across major IDEs
AWS's agentic, spec-driven AI IDE that turns prompts into requirements, designs, and sequenced tasks, then builds them with parallel agents
Google's asynchronous AI coding agent that clones your repo, plans changes, runs in a cloud VM, and opens a reviewable pull request
Yes, for individuals. Antigravity has been free since its November 2025 launch and remains at $0 for individual use during the public preview. It is metered by a usage quota rather than a dollar cap, refreshing on a rolling five-hour cycle up to a weekly ceiling. Google has reduced those free limits several times since launch.
Not standalone. Higher quotas are bundled into Google AI subscriptions: AI Pro at $19.99/month, AI Ultra from $99.99/month with roughly 5x the Pro Antigravity limit, and a $200/month top Ultra tier with roughly 20x. Pro and Ultra subscribers can also buy pay-as-you-go top-up AI credits at $25 per 2,500 credits.
It launched on Gemini 3 Pro plus Anthropic's Claude Sonnet 4.5 and OpenAI's open-weights GPT-OSS. As of mid-2026 the picker covers Gemini 3 Pro and Flash, Claude Sonnet and Opus, and GPT-OSS-120B. The list is curated through Vertex Model Garden — there is no bring-your-own-key or local model support.
For code generation and reasoning quality, Claude remains the stronger choice, which is why Antigravity itself ships Claude models in its picker. Antigravity's advantages are elsewhere: it is free for individuals, and its agent manager orchestrates multiple sandboxed agents in parallel. Many developers use both — Antigravity as the harness, Claude as the model doing the hard thinking.
Google replaced it. Antigravity CLI launched publicly on 19 May 2026 as a Go rewrite, and Gemini CLI stopped serving requests on 18 June 2026 for free users and Google AI Pro and Ultra subscribers. Organisations on Gemini Code Assist Standard or Enterprise licences kept their existing access unchanged.
Partly. The original Antigravity IDE launched as a modified VS Code fork, so extensions and keybindings largely carry over, and it is still shipping. Antigravity 2.0, introduced at Google I/O 2026, is a separate standalone desktop app built around agent orchestration rather than a VS Code fork. Both run on the same agent harness, alongside the CLI and SDK.
Yes. The Antigravity SDK gives programmatic access to the agent harness so you can define custom agents and host them on your own infrastructure. Separately, the Antigravity Agent is available in preview through the Gemini API Interactions API and Google AI Studio, running in a Google-hosted Linux sandbox and billed pay-as-you-go on underlying model tokens.
There is no announcement that it will, and Google is actively investing — it consolidated Gemini CLI into Antigravity and shipped version 2.0 at I/O 2026, which signals commitment rather than abandonment. That said, Google retired Gemini CLI roughly seven months after Antigravity launched, so treat any preview-stage Google developer product as carrying real discontinuation and repricing risk before you build a team workflow on it.