What Is Google’s Antigravity? Everything We Know About the New Agentic Platform

Google has launched Antigravity, a new agentic development platform powered by Gemini 3 and other leading AI models. The experimental tool is now available in free public preview for Windows, Mac and Linux. The move marks Google’s latest push into next-generation coding tools designed for higher-level, task-oriented development.

The debut of Antigravity has drawn major attention across the developer world. Google positions the platform as a shift toward agent-first software creation, where AI models can manage tasks, verify code and deliver actionable outputs. The company outlined the tool during a press briefing, noting that Antigravity aims to redefine how developers interact with AI-powered coding systems.

How Google’s Antigravity Works and What Makes It Different

Antigravity allows developers to choose between multiple models, including Google’s Gemini 3, Anthropic’s Claude Sonnet 4.5 and OpenAI’s GPT-OSS. Google says rate limits refresh every five hours, giving developers consistent access during the preview period. This flexibility makes Antigravity one of the first IDE-style tools to offer model switching at launch.

The platform goes beyond typical AI code assistants. Instead of displaying raw tool calls or only showing the final output, Antigravity introduces “artifacts.” These are structured deliverables such as task lists, implementation plans, screenshots and browser recordings. Google says this improves trust by letting users validate each step of the agent’s work. According to information reported by major outlets including CNN and BBC News, Google has increasingly emphasized verification as a key part of its AI development strategy in 2025.

Developers can also steer the agent using Google Docs-style comments on generated artifacts or by responding to screenshots. Over time, Antigravity’s internal knowledge base adapts, storing code snippets and successful workflows. This allows the agent to improve its future performance without starting from scratch each session.

Another key feature is the “agent-first Manager surface.” This interface acts as Mission Control for supervising, spawning and orchestrating agents. Instead of embedding agents inside each view, Antigravity embeds the views inside the agent. Google says this reverses the traditional IDE hierarchy and reflects how high-capacity models like Gemini 3 operate across multiple surfaces at once.

The platform can run and test applications inside an integrated Chrome browser. When tasks are complete, the agent produces a full walkthrough to show how the final product behaves. Short in-editor completions, in-line commands and real-time coding assist functions round out the environment. Antigravity joins Google’s broader ecosystem of agentic tools, including Gemini CLI, Jules and vibe coding in AI Studio.

The Bigger Picture for Developers and the Future of Agentic Coding

Antigravity arrives at a time when agent-based coding continues to advance rapidly. Tools from major companies have moved far beyond autocomplete and now support autonomous task execution, planning and debugging. Google’s approach suggests a shift toward transparency and multi-model flexibility as core features rather than add-ons.

Long-term, the platform could influence how developers design workflows, collaborate with AI and manage complex projects. Its agent-first design may encourage more modular and parallel task execution. Its artifact system may also set standards for auditability in AI-generated code.

Google’s Antigravity marks a major step in agentic development. As adoption grows, the platform may help shape how AI-powered coding evolves in 2025 and beyond, especially as developers explore what an agent-first environment can deliver.

FYI (keeping you in the loop)-

Q1: What is Google’s Antigravity?

Antigravity is a new agentic development platform from Google. It combines Gemini 3 with other leading AI models to generate, test and verify code. It operates through artifacts and agent-driven workflows.

Q2: Is Google’s Antigravity free?

Yes. Antigravity is available in free public preview with generous usage limits. Google says rate limits refresh every five hours during the preview period.

Q3: Which AI models does Antigravity support?

Developers can choose Gemini 3, Claude Sonnet 4.5 and OpenAI’s GPT-OSS models. Google says model access depends on capacity but refreshes regularly.

Q4: How does Antigravity verify the code it generates?

The system uses artifacts such as task lists, screenshots and browser recordings. It also runs applications inside a Chrome browser to confirm behavior before presenting a walkthrough.

Q5: How does Antigravity compare to Gemini CLI and Jules?

Antigravity focuses on agent-first workflows. Gemini CLI remains Google’s most widely used coding tool, while Jules and Antigravity operate as experimental platforms.