JetBrains has added Gemini 3 Pro as a selectable model in AI Chat and Junie. For engineering leads this is less about a shiny model and more about predictable behaviour, safe rollout and measurable gains in delivery. Below is our take as a JetBrains Premium Reseller – what JetBrains released, where it can help, and how to enable it with the right guardrails.

Source note: all capability claims are taken from JetBrains’ announcement and demo materials. We have not independently tested Gemini 3 Pro.

What JetBrains released

Gemini 3 Pro now powers AI Chat and can be chosen for Junie. JetBrains highlights improved reasoning, closer adherence to instructions and stronger multimodal performance, with particular gains in frontend and long-context tasks. The goal is to move from prompts to production-ready changes with fewer edits during review.

Where this can help

Teams that work with long prompts, mixed inputs or UI scaffolding stand to benefit first. Gemini 3 Pro’s instruction following should make refactors and guided edits easier to steer, while Junie’s agent workflow can plan multi-step changes that you approve in stages. Used well, this shifts effort from manual boilerplate and iteration to review and refinement.

Picking the model and setting guardrails

Keep the default Auto selection for day-to-day use, and pin Gemini 3 Pro when you expect longer context, UI generation or broader reasoning. Before rollout, agree on scope – what the model may read, where it may write and who approves agent actions. Treat AI output as authored code subject to your usual tests, linters and quality gates. Keep prompts and diffs visible in pull requests so reviewers see both the intent and the change.

Enablement in practice

Licences come first – JetBrains AI is required and trials can be started from within the IDE. Make sure the Junie plugin is current across the pilot group. Begin with a small cohort and publish a one-page playbook that sets expectations for prompts, commit messages, revert strategy and code-review etiquette. Isolate agent-generated work on branches, keep pull requests small and reviewable, and use your CI to enforce standards. For privacy, restrict sensitive repositories and confirm the JetBrains AI data-handling settings that fit your organisation.

Risks to watch

Over-automation can creep in when agents are left to make sweeping edits. Keep tasks narrow, require named ownership on pull requests and pair-review higher-impact changes. Style drift is another common issue – lean on formatters, templates and project conventions to keep output consistent. If scope expands, pause and re-plan rather than letting a large chain of edits land at once.

How to enable Gemini 3 Pro

In AI Chat, start a new conversation – the model uses Auto by default. In Junie, open Settings → Junie → Models and select Gemini 3 Pro. Activation requires a JetBrains AI subscription; if you do not have one, start a trial from the JetBrains AI widget in the IDE. For step-by-step setup, follow JetBrains’ official guide.