So gemini 3.1 pro launched while 3.0 models seem to still be in preview. Many of us in europe are very eager to specify some european location rather than just “global” when using these models through Vertex AI.
Are any from Google/Vertex/etc able to comment on a timeline or expectations here re. EU availability? Previously, 2.5 pro/flash were made available in EU locations no more than a couple of months after the announcement but now it seems to drag out…?
For many EU businesses this is the deciding question with regards to using gemini or GPT/Claude models instead in a production environment.
I support this very much. For previous generations, the time gap between the global-preview release and the release in specific locations was much shorter than for gemini 3. The pro version was released in November, a 3.1 generation exists already and we have not received yet in EU. Can anyone shed some light on when are we getting it or why are we not getting it?
Same here, we are currently seeing ourselves forced to move our production deployment from Gemini 2.5 Pro to Claude Sonnet/Opus 4.6 because Gemini 3.x models are still not available on Europe clusters. Google is really fumbling this one
Just received an email prompting me to migrate to Gemini 3 due to a pending discontinuation of Gemini 2.5 sometime after October 16. It is a bit pathetic to send this out now given the unavailability of the newer models in europe
Similar to TimMathis, just received an email for 2.5 discontinuation. We will have a major issue for our customers if 2.5 is discontinued but 3 is not available in Europe only. We are actively migrating our clients away from Google solutions because of this.
More than that, we definitely need a guarantee with 3.x that data processing - not just data at rest - is done in the EU.
We can’t afford to wait un unknown amount of time for google to GA 3.x in EU data centers, so we’ll be keeping our eyes open for alternatives to google now.
After waiting for months we have finally migrated to OpenAI Enterprise. Google went from having the best EU-based AI processing to practically abandon it. Not only that but they don’t offer serious high-availability at all (priority/premium pay-as-you go only for US/global). Their provisioned throughput is flat and expensive meaning you pay 10 times as much as pay-as-you-go yet still get 429 (too many requests) in peak hours. All of the time. Really sad this happened.
Adding our voice to this thread. We’re in the same position.
We use Gemini for video understanding (best-in-class for our use case), but we have two blockers that make migration non-trivial right now:
EU data residency regulations: we can’t process data outside of Europe. Full stop.
Data co-location at high load: running at scale means we need inference and data to live in the same region. Cross-region setups are cost-prohibitive for us.
Honestly, the thing that stings most isn’t the deprecation itself but it’s feeling left without a viable path. A deadline with no roadmap for EU-based customers puts us in a difficult position.
If anyone from Google Cloud is reading this: we’d genuinely appreciate more proactive communication for customers with residency requirements.
As mentioned by everyone else, it would be nice to have the option for EU deployments, we have multiple agentic use cases that require EU data residency to be viable, and we’ve already seen positive benchmarks using Gemini models for this purpose.
We’re also waiting for european locations for gemini 3 models. The deprecations and shutdown dates is very contradictory and it’s not clear when the regional locations will be available.
Same here, now the page that was describing the regional location of the models has been discontinued (Data residency | Generative AI on Vertex AI | Google Cloud Documentation) so there is absolutely no information on this matter. Please make this key information visible to everyone. Otherwise I understand that developers like us will need to migrate to other LLMs providers
That is really good news but do you know if this means that VPC-SC perimeter’s are well bounded in the EU? I am working on a client project and have set the infrastructure to a specific EU-subregion but it doesn’t look like the Gemini 3.X model family will be made available on that granular a level. I tested to check if 3.5 was available in different sub-regions but it wasn’t.
We’ve got another email that states gemini 2.5 pro will be deprecated in October. But there is no Pro substitute for it in Europe? This means degradation on bigger nuanced tasks.
What is going on with these deprecations? Shouldn’t there be at least a proper substitute before depreciating certain models?
It’s already troublesome to see that we have to work with a Pro model that is as old as 2.5 currently is. Really hope Google can fix this.
