Gemini Enterprise switches on its new model by default on 27 October. Claude now runs in the EU region too.

Google's release notes for Gemini Enterprise carry two changes from the past week that land on an admin's desk. On 13 October Google made its newest model, Gemini 4 Argon, generally available in the Gemini Enterprise web app, the mobile app and the AI developer tools. In the app it arrived switched off. From 27 October the toggle defaults to on. On 15 October Google added a second change: admins can now offer Anthropic's Claude Opus 5.5 and Claude Sonnet 5.5 inside the same app.

If your company does not pay for Gemini Enterprise, none of this applies to you. If it does, someone needs to look at the settings before next Tuesday.

Gemini 4 Argon: off now, on from 27 October

Google's documentation lists Argon as turned off in the global, US and EU locations, with the default switching to on from 27 October 2026. In the EU multi-region Google says both data at rest and machine learning processing stay in the region. The same model reached Workflow Builder and workflow agents on 14 October under the same 27 October date. In the AI developer tools it is already on for licensed users who keep the default model selection.

A new default model changes answers that people have learned to rely on. Prompts that worked last week may give longer, shorter or differently structured output. Workflows built on the old model may behave differently too. That is no reason to block the upgrade, but it is a reason to test it first.

Claude inside Gemini Enterprise

Claude Opus 5.5 and Sonnet 5.5 are off by default. An admin with the Gemini Enterprise Admin role turns them on under Feature Management, in a section called Enable third-party models, after enabling billing and accepting the terms of service. Users then pick them from the model selector. Google's setup page says the first activation can take up to 15 minutes.

The cost works differently from the seat licence. Google bills third-party models per token, pay as you go, and the project needs pay-as-you-go usage above quota switched on. Google itself recommends setting a monthly spend limit. The models are available in the global location and the US and EU multi-regions, and Google's locations page lists at-rest data residency and machine learning processing for both. They are not offered in the Starter edition or in single-country regions.

What to do on Monday

• ✳Find out who in your company holds the Gemini Enterprise Admin role. If the answer is nobody in particular, fix that first
• ✳Before 27 October, decide whether Argon should switch on for everyone, for a test group first, or stay off for now. Doing nothing means it switches on
• ✳Run your five most used prompts and any Workflow Builder flows through Argon and compare the results with what people get today
• ✳Check that your Gemini Enterprise app runs in the EU multi-region if data residency matters to you. Both Argon and the Claude models keep data and processing in the EU there
• ✳Turn Claude on only with a spend limit already set, and decide who gets access. Per-token billing does not show up in the seat price your finance team already knows

Two settings changes this small rarely make news. They matter because one of them happens automatically and the other adds a bill that grows with use. An admin who spends an hour on both this week avoids a confused helpdesk and an unexpected invoice next month.