Gemini 3.5 Flash removed from the global-region endpoint in Gemini Enterprise from 4 August 2026
From 4 August 2026, gemini-3.5-flash is no longer available in the global-region endpoint of the Gemini Enterprise app. Here's what to check.
From 4 August 2026, gemini-3.5-flash is no longer listed as an available model in the global region endpoint of the Gemini Enterprise web app. If your organisation routes Gemini Enterprise traffic through the global region, this is worth checking now rather than finding out when a workflow breaks.
This is a scoped change. It affects the in-app model selector within the Gemini Enterprise web app and Agent Designer, specifically for teams using the global multi-region endpoint. It is not a full retirement of the gemini-3.5-flash model ID across all of Google Cloud, and the model itself has a stated minimum lifecycle of 12 months from its general availability release on 19 May 2026.
What actually changed, and when
The August 4 removal is the second step in a two-part process that started in June.
From 8 June 2026, Google removed the feature management toggle that allowed admins to turn gemini-3.5-flash on or off for their users. From that point, the model was enabled by default and could not be disabled. That toggle removal applied across all three multi-regions: global, US, and EU.
Then, from 4 August 2026, the model has been pulled from the global region entirely. Teams on the US or EU multi-region endpoints are not affected by this specific August change.
Why this is happening
Google released gemini-3.6-flash in late July 2026, making it available in the global and US multi-regions. The newer model improves on token efficiency, code generation, and agentic planning, and is priced lower than gemini-3.5-flash. The removal of the older model from the global region follows a pattern of consolidating the model lineup as newer options become available, rather than maintaining every model version across every endpoint indefinitely.
It is worth noting that gemini-3.5-flash remains available in the EU multi-region, which is relevant for organisations with GDPR or data residency requirements that confine processing to EU geography.
What you need to check
If your team uses the Gemini Enterprise web app or Agent Designer and you have not thought about which region endpoint you are on, now is the time to confirm.
Check your region configuration. If you are routing traffic through the global endpoint, gemini-3.5-flash will no longer appear as a selectable model. Any saved configurations, agents, or workflows that referenced that model in the global region will need to be updated.
Review your model selector settings. Admins can manage which models are available to users through the Google Cloud console under the Gemini Enterprise feature management section. The model availability controls require the “Enable model selector” toggle to be switched on before individual models can be configured. Google’s manage web app features documentation walks through this.
Consider moving to Gemini 3.6 Flash. If your use case was well-served by gemini-3.5-flash, gemini-3.6-flash is the natural next step and is already available in the global and US multi-regions. One thing to test before switching: the newer Flash models do not support custom values for temperature, top-K, top-P, frequency penalty, or presence penalty. If your Agent Designer configurations set any of those parameters, they will be silently ignored or, in some cases, return an API error. Check your agent configurations before rolling out broadly.
EU multi-region users can hold. If your organisation deliberately routes through the EU multi-region for compliance reasons, gemini-3.5-flash remains available there. No immediate action is required, though it is sensible to plan for eventual consolidation on that endpoint too.
A note on the earlier admin toggle removal
One thing that caught some teams off guard between June and August: because the admin toggle was removed on 8 June 2026, organisations that had previously set a specific policy around gemini-3.5-flash availability lost the ability to enforce it. The model was switched on by default across all regions from that date. If you are reviewing your model governance posture now, it is worth checking whether any user-facing changes happened without an explicit admin decision on your end.
The broader picture
This change is a good reminder that model availability in enterprise AI products is not static. Google is iterating quickly on the Gemini model family, and the endpoint and region configuration that worked at rollout may not match what is available six months later. Building a habit of reviewing model configuration settings quarterly, particularly after major release notes updates, will save time compared to triaging unexpected behaviour in production.
The Gemini Enterprise release notes are the authoritative place to track these changes. They are updated regularly and document not just new features but removals and deprecations, which are often buried below the headline announcements.