Google Releases Free AI-Powered IDE: Antigravity
Google has entered the AI coding tools arena with Antigravity, a free IDE that comes bundled with Gemini 3 Pro, Claude Sonnet 4.5, and GPT-OSS models.
What You Get for Free
- Access to Gemini 3 Pro, Claude Sonnet 4.5, and GPT-OSS
- Unlimited tab completions
- Unlimited command requests
- Available for Mac, Windows, and Linux
The Backstory
Antigravity is built by the Windsurf team, which Google acquired for $2.4 billion earlier this year. It's essentially a VS Code fork with an "agent-first" approach, meaning you can delegate entire coding tasks to AI agents rather than just getting autocomplete suggestions.
The Killer Feature: Manager Mode
This is where it gets interesting. Antigravity introduces two ways of working:
Editor Mode is the familiar experience. You type, you get suggestions, you have an AI chat panel on the side. Standard stuff.
Manager Mode is the game changer. Think of it as Mission Control for AI agents. You assign a high-level goal like "Refactor the auth middleware to support JWT rotation" and an autonomous agent takes over. It plans the steps, edits files, runs terminal commands, and here's the crazy part - it spins up a browser to physically click through your UI to verify the fix actually works.
You can even run multiple agents in parallel. One agent fixing a CSS bug on the frontend while another writes unit tests for the backend. That's something Cursor don't offer yet.
Why Is It Free?
Let's be honest, it's not as polished as Cursor or Claude Code yet. Early users report hitting rate limits after about 20 minutes of intensive use, and there are occasional "model provider overload" errors. Google needs real-world usage data to fine-tune the experience.
The free tier likely won't last forever. Based on similar launches, expect 3-6 months before a paid model kicks in (probably $20-40/month).
Benchmark Battle: Gemini 3 vs Claude Sonnet 4.5
For those who care about numbers:
| Benchmark | Gemini 3 Pro | Claude Sonnet 4.5 |
|---|---|---|
| LiveCodeBench Pro | 2,439 | 1,418 |
| SWE-Bench Verified | 76.2% | 77.2% |
Gemini 3 gives more complete responses, while Claude tends to be more concise. Both are excellent for coding, just different styles.
Should You Try It?
If you're already happy with Cursor or Claude Code, there's no rush to switch. But if you want to experiment with multiple AI models without paying anything, Antigravity is worth a look.
Download it at antigravity.google