When Google launched Antigravity in November 2025, it looked like a me-too IDE — a free, Gemini-powered answer to Cursor. The 2.0 release unveiled at I/O 2026 is a different animal entirely. Google isn't selling an editor anymore. It's selling an agent control plane.
What Changed in 2.0
The original Antigravity was an agent-first IDE: one editor, one assistant, the familiar Cursor template. Version 2.0 explodes that into a platform with four front doors:
- A revamped desktop app built around multi-agent orchestration — the new centerpiece
- A new CLI written in Go, which Google positions as faster and more responsive than its predecessor
- An SDK for building custom agents on Google's coding infrastructure
- Deep native integrations with Google AI Studio, Firebase, and Android
The headline capability is parallelism. In the 2.0 desktop app you can set several agents working on a problem simultaneously, design custom subagent workflows, and schedule tasks that run in the background without you babysitting them. This is the shift Google is betting on: the developer stops being a coder-with-an-assistant and becomes a manager of a pool of AI workers.
If your day already looks like "kick off three agents, review their PRs, merge the good one," Antigravity 2.0 is the first mainstream tool designed for that exact loop.
The Engine: Gemini 3.5 Flash
Antigravity 2.0 runs on Gemini 3.5 Flash, the model Google unveiled at the same I/O. Google's claim is aggressive: 3.5 Flash outperforms Gemini 3.1 Pro on coding and agentic benchmarks while running roughly four times faster than competing frontier models.
There's a tidy bit of dogfooding here — Google says Gemini 3.5 Flash was itself co-developed using Antigravity. Take vendor benchmark claims with the usual skepticism, but the strategic logic is sound: when you're orchestrating many agents at once, raw per-token speed matters more than squeezing out the last few points of single-shot quality. A fast, cheap, "good enough" model you can run ten copies of often beats one slow genius.
The desktop app also gains native voice commands, letting you talk to your agents instead of typing every instruction.
The Pricing Shuffle
Google reworked its paid tiers around the launch:
- AI Pro remains the entry paid plan at roughly $20/month.
- A new $100/month tier lands in the middle, offering about 5× the usage limits of AI Pro.
- The top tier dropped from $250 to $200/month, delivering around 20× the Pro limits.
That $100 price point is no accident. It plants Google directly opposite OpenAI's ChatGPT Pro and Anthropic's Claude Max, which both sit at the same $100/month mark for heavy coding use. The $200 top tier matches Anthropic's most expensive Claude Max plan. Google is fighting for developer loyalty on price as much as capability.
The Casualty: Gemini CLI
Platforms this ambitious leave bodies behind, and 2.0 has one: Gemini CLI is being retired. Google confirmed that consumer access to Gemini CLI and the Gemini Code Assist IDE extensions ends June 18, 2026 for AI Pro, AI Ultra, and free-tier users.
Enterprise customers on Gemini Code Assist Standard or Enterprise licenses keep their access, so this isn't a total kill — but the message to individual developers is unambiguous. The Go-based Antigravity CLI is the designated successor, and if you built workflows on Gemini CLI, the migration clock has already run out.
What's Genuinely Good
- Orchestration is first-class, not bolted on. Most "agentic" tools still center a single chat thread. Antigravity 2.0 treats running many agents as the default mental model.
- The AI Studio → local export path is a real workflow win. Prototype in the browser, then pull the project — with full context — into your local instance.
- A Go CLI plus an SDK means Antigravity reaches into terminals and CI pipelines, not just the editor window.
- Aggressive pricing undercuts the obvious alternatives for high-volume users.
What's Missing or Worrying
No tool this young is a clean win. A few honest caveats:
- Lock-in is the whole strategy. Native ties to AI Studio, Firebase, Android, and Gemini are a feature and a trap. The more of your workflow lives here, the harder Google is to leave — and the company just demonstrated, by sunsetting Gemini CLI, that it will retire tools on its own schedule.
- Benchmark claims are Google's own. "Beats 3.1 Pro, 4× faster" is a vendor statement measured on vendor-chosen tasks. Treat it as a hypothesis until independent evaluations land.
- Orchestration raises the stakes on review. Running five agents in parallel multiplies output — and the volume of plausible-looking code you now have to vet. Multi-agent speed is only a win if your review process can keep up.
- A churn-prone roadmap. A platform that rebranded, re-architected, and killed its own CLI inside seven months is moving fast. That's exciting if you like the bleeding edge, exhausting if you need stability.
The Bottom Line
Antigravity 2.0 is the clearest bet yet that the future of coding is orchestration, not autocomplete. If your work has already shifted from "AI helps me write code" to "I supervise a swarm of agents," it's the first tool built for that reality from the ground up — and the pricing is sharp enough to take seriously. Just go in clear-eyed: you're not adopting an editor, you're joining an ecosystem, and Google has already shown it'll move the furniture whenever it suits the platform.
