Gemini 4 Argon is coming to Gemini Enterprise by default on 27 October 2026
Gemini 4 Argon launched in Gemini Enterprise on 13 October with its toggle off. From 27 October it becomes the default model for all users.
Google’s newest frontier model, Gemini 4 Argon, became generally available in Gemini Enterprise on 13 October 2026, arriving in the web app, mobile app, and Workflow Builder with its toggle switched off by default. That changes on 27 October 2026, when Google will flip the default to on for all users. Admins have a narrow window to decide whether to act before that happens.
What is Gemini 4 Argon?
Gemini 4 Argon (gemini-4-argon) is Google’s first frontier model release in more than seven months, and it marks a naming shift: rather than continuing the Gemini 3.x line, Google has jumped to a new naming scheme with Argon as the first member of the Gemini 4 family.
The model is designed for deep, multi-step work across complex domains: coding, legal, finance, cybersecurity, and science. Google reports it ranks first on the Vals Index, which benchmarks model performance across finance, coding, legal, and tax tasks. On CWE-bench v1, a vulnerability remediation benchmark, it ties for first place at 68%. The output token limit has been raised to 1 million tokens, matching the input context window, which is a significant increase from the previous 64,000-token output cap.
The toggle timeline admins need to know
Here is the sequence of changes in the Gemini Enterprise apps:
- 13 October 2026: Gemini 4 Argon becomes generally available in the Gemini Enterprise web app and mobile app. Toggle is off by default.
- 14 October 2026: Gemini 4 Argon becomes generally available in Workflow Builder, also off by default.
- 27 October 2026: The toggle defaults to on for all users in both the web and mobile apps, and in Workflow Builder.
That gives admins roughly two weeks from the GA launch date to review their configuration before the model becomes active for everyone in their organisation.
For AI developer tools, the behaviour is slightly different. Gemini 4 Argon is enabled by default for licensed users when Antigravity authorised models is set to the default selection. If your organisation has specified a custom model list, an administrator must manually add Gemini 4 Argon. It will not appear automatically.
What admins should do before 27 October
The main things to check before the default switches on:
Review your model list. If your organisation uses a custom model list in AI developer tools, log into the Feature Management tab in the Gemini Enterprise page of the Google Cloud console and manually add gemini-4-argon if you want users to have access.
Check your spending controls. Gemini Enterprise supports hard spend limits per project via the Cloud Billing Console. Gemini 4 Argon is priced at $2 per million input tokens and $10 per million output tokens during the introductory period, rising to $4 and $20 thereafter. Cached input is priced at 95% below the standard input rate. If your existing spend caps were calibrated around Gemini 3.x usage, it is worth revisiting them before a more capable and potentially more heavily used model becomes the default.
Consider whether you want to delay. If your organisation is not ready to roll out Gemini 4 Argon on 27 October, you can manually switch the toggle off in the Feature Management tab before that date. This will prevent it from being automatically enabled for your users.
A few other things in this release worth noting
Workflow Builder has gained connections to Microsoft Teams, Notion, and Salesforce, allowing workflows to search data and automate actions across those third-party apps.
From 15 October 2026, administrators can also make Anthropic’s Claude Opus 5.5 and Claude Sonnet 5.5 available to users in the Gemini Enterprise app. Once an admin enables a model in the Enable third-party models section of the Feature Management tab, users can select it from the model selector. Claude models are available in the global, us, and eu regions and are charged on a pay-as-you-go basis.
Administrators can now configure citation_mappings for custom Model Context Protocol (MCP) data stores via the REST API, allowing source citation chips to appear in responses. This is generally available as of 16 October 2026.
The broader picture
Gemini 4 Argon first appeared on 30 September 2026 in a limited rollout to cyber defenders through Google’s Fairwind Programme. The general availability launch on 13 October opened it up to Google AI Pro and Ultra subscribers globally, and simultaneously brought it into Gemini Enterprise.
For most Gemini Enterprise organisations, the practical question is not whether to use the model eventually, but whether to be ready for it on 27 October or to take a more controlled approach. Given the output token increase alone (from 64,000 to 1 million tokens), prompts and agents that previously hit limits may behave differently, and that is worth testing before the switch flips automatically.
The Feature Management tab in the Google Cloud console is where all of these controls live. If you have not been there recently, now is a good time to look.