Building with Gemini 3 in Jules

On Tuesday we introduced Gemini 3, Google’s most intelligent model that can help bring any idea to life. Today, we’re excited to share that Gemini 3 Pro will start to be available in Jules, our always-on, autonomous coding agent, for Google AI subscribers. It is rolling out to Google AI Ultra subscribers starting today, and will arrive for Pro plan subscribers in the coming days.

Gemini 3 surpasses 2.5 Pro at coding, mastering both agentic workflows and complex zero-shot tasks, and serves as a new foundation of intelligence for what’s possible with an agentic coding model. So what does that mean for Jules?

Get the most out of Gemini 3 in Jules

Gemini 3 Pro’s improved agentic capabilities in Jules translate to clearer reasoning, stronger intent alignment, and a noticeable lift in day-to-day reliability. Tasks that span several steps hold together more naturally. Plans stay coherent and the system maintains context with less effort from you.

Always on, ready to build

We’ve been working hard over the last two months to bring Jules to more surfaces, like the terminal with Jules Tools, or the Jules extension in Gemini CLI, or wherever you want Jules to surface via the Jules API. What’s cool about Gemini 3 in Jules for Ultra and Pro users is that all of these surfaces have the same view of your project so you can move between them without losing context.

This is critical because we believe a good developer tool creates more space to think. It should be there when you reach for it, no matter where you are working, and it should be able to handle the volume and complexity of real software projects. That is the direction we have been pushing Jules. The goal is to create an agent that is always on, available everywhere you work and able to follow through on multi step tasks without you having to manage every transition. Our goal is to make the work less about managing the tool and more about shaping your ideas.

Other recent improvements in Jules

We have been busy working to make everyday use of Jules feel more predictable and transparent. In the past few months we’ve shipped some exciting updates available to all Jules users:

• The CLI now supports parallel runs, works cleanly on Windows, has a simpler invocation flow, and includes a diff viewer so you can confirm changes before applying them.
• The Jules API is stable and ready for custom workflows, automation, and deeper integrations.
• We added support for environment variables and improved how memory carries through a session.
• The critic agent now steps in more reliably and re-engages after replanning, which helps the system stay on track during longer tasks.
• We have also been cleaning up the rough edges. Git handling is safer and less surprising.
• VM performance is improving, which shortens the time between a request and a meaningful response.

And we have more on the way:

• Attach a directory without having to go through GitHub so you can start quickly without worrying about where the code lives.
• Titles will update automatically to keep your history readable.
• The web UI is gaining CLI shortcuts for faster actions.
• Start modes are being simplified.
• Automatic PR creation is in progress for experimental workflows.

Check the changelog for all of this and smaller updates like verification, additional CLI work, and more critic against and quality improvements.

We hope you enjoy using Gemini 3 in Jules. To get started, visit jules.google.com.