In a broadside to Microsoft on the opening day of Ignite 2025, Google announced the release of Gemini 3, its most advanced AI model yet.
“Today we are introducing Gemini 3, our most intelligent model that can help bring any idea to life,” Google’s Logan Kilpatrick explains. “Built on a foundation of state-of-the-art reasoning, Gemini 3 Pro delivers unparalleled results across every major AI benchmark compared to previous versions. It also surpasses 2.5 Pro at coding, mastering both agentic workflows and complex zero-shot tasks.”
Google has been a hot streak in AI lately, triggered by its “Nano Banana” image-to-video and generative image AI models. Gemini 3 seems poised to continue the success, with Google claiming massive performance and accuracy gains over its predecessors and competition like OpenAI ChatGPT 5.1 and Claude Sonnet 5.
Gemini 3 Pro is also an easy upgrade, Google says, noting that it can “fit right into existing production agent and coding workflows” while enabling all-new use cases too. It’s available now in preview through the Gemini API in Google AI Studio and Vertex AI for enterprises, and in Google AI Studio for everyone.
Also new, Google is introducing its Antigravity agentic development platform as a showcase of what’s possible with Gemini 3. This looks like yet another Visual Studio Code-based IDE (integrated development environment) but with agents for planning and automating tasks. The Windows version of this IDE is available in x64 and Arm versions, too.