Google postpones removal of Gemini 3.5 Flash from global region in Gemini Enterprise app
Google has reversed its 4 August 2026 deadline to remove Gemini 3.5 Flash from the global-region endpoint. No new date has been set.
On 5 August 2026, Google published a correction to its Gemini Enterprise release notes reversing a deadline that had been set just two weeks earlier. The original 21 July 2026 release note stated that Gemini 3.5 Flash would be removed as a model from the global region in the Gemini Enterprise app on 4 August 2026. That note now carries a correction flag, and the August deadline no longer stands. Google says the removal has been postponed and that a revised schedule will be provided when available.
No new date has been announced.
What the original announcement said
The 21 July release note was specific: Gemini 3.5 Flash would be removed from the model selector in the global-region endpoint of the Gemini Enterprise app on 4 August 2026. This was not a full model retirement. Google’s model lifecycle documentation lists gemini-3.5-flash as having been released on 19 May 2026 with a retirement horizon of at least 12 months from that date. The July announcement was about endpoint availability, not the model disappearing from Google Cloud altogether.
That distinction matters, but it did not make the original deadline any less disruptive for organisations that had standardised on the global endpoint. Removing a model from a regional endpoint can affect latency characteristics, SLO calculations, and any tooling that references a specific region. For some teams, two weeks is not a comfortable migration window.
Why this matters for your organisation
If your organisation was preparing to move workloads away from the global endpoint ahead of 4 August, you can pause that work. The pressure of the original deadline is gone.
That said, the removal is still coming. Google has not said it will keep Gemini 3.5 Flash in the global region indefinitely, only that the 4 August date no longer applies and a new schedule will follow. The practical advice is to bookmark the Gemini Enterprise release notes page and check it regularly. When a revised date appears, the migration planning you may have already started will serve you well.
It is also worth noting that supervised fine-tuning for gemini-3.5-flash is already restricted to us-central1 and europe-west4, and tuned model serving is restricted to the US and EU multi-region endpoints. Those constraints are unrelated to the global-region removal and remain in place.
A tight schedule that caught teams off guard
The original 21 July announcement gave organisations less than two weeks to react before the 4 August deadline. That timing sat alongside a busy period of Gemini Enterprise changes: Gemini 3.6 Flash had been made available in the global and US multi-regions shortly before the July note appeared, the Gemini 3.5 Flash feature management toggle had been removed in June 2026 across the Global, US, and EU multi-regions, and NotebookLM Enterprise had been renamed to Gemini Notebook Enterprise during the same period.
For IT administrators tracking these changes, the correction on 5 August is a relief, but it also illustrates a broader point about following Google Cloud release notes closely. Endpoint-level changes can arrive with short lead times, and the correction mechanism, while welcome, is not a substitute for catching the original announcement early.
What else changed on 5 August
The 5 August release note that carried the Gemini 3.5 Flash postponement also included a useful simplification for teams using custom MCP servers with Gemini Enterprise. Description fields are no longer required when setting up a custom MCP server. Gemini Enterprise now automatically calls a server’s tools/list endpoint to discover tools, parameter schemas, and capabilities directly, using those explicit definitions to route requests rather than relying on description text. Existing custom MCP data stores do not require any updates as a result of this change.
The same release notes period also introduced two new trace spans for Gemini Enterprise’s data connector workflow: execute_tool, representing tool execution on the agent orchestration layer, and invoke_connector, representing request logic on the connector execution layer. These give teams better visibility into end-to-end parent-child workflows when debugging or auditing connector behaviour.
Separately, the Gemini Enterprise mobile app reached general availability for organisations using third-party identity providers, and can now be connected without being on an allowlist.
What to do now
There is no immediate action required on the Gemini 3.5 Flash global-region question. If you were mid-way through a migration plan triggered by the 4 August deadline, document where you got to. When the revised schedule appears, you will be better positioned than organisations starting from scratch.
Keep an eye on the Gemini Enterprise release notes directly rather than relying on third-party summaries for time-sensitive changes like this. The correction appeared there first, and given the speed at which these notes have been moving through mid-2026, that is where the new date will appear too.