Gemini 3.5 Pro Was Supposed to Ship in June. It’s June 19, and Only Big Companies Can Touch It.

Let me save you some time. If you have been holding your AI plans waiting for Gemini 3.5 Pro, stop holding. As of today, you still cannot get it.

Here is the situation. Back at Google I/O on May 19, Sundar Pichai teased it from the stage. His words: "We’re also hard at work on 3.5 Pro. It’s already being used internally, and we look forward to rolling it out next month" (TechTimes). "Next month" meant June. The crowd reportedly groaned — they wanted it then.

It is June 19. The model still has not shipped to the public. As of this writing, Gemini 3.5 Pro remains in limited preview for select Vertex AI enterprise customers only. It is not in the consumer Gemini app, not in Google AI Studio, not in the paid Gemini subscription (Codersera). If you are a normal person or a small business, you cannot use it. If you are a big enterprise with a Vertex AI relationship, maybe you can.

What it is supposed to be

On paper, it is a monster. A 2-million-token context window — the largest of any production frontier model right now (Growwing Assistant). A "Deep Think" reasoning mode. Full multimodal understanding across text, images, and more. Expected pricing in the neighborhood of $15 per million input tokens and $60 per million output (Growwing Assistant). This is the model meant to absorb everything Google used to route to Gemini Ultra.

That 2M context number is the headline for a reason. For anyone working with giant documents, whole codebases, or long meeting transcripts, context size is not a spec-sheet brag — it is the difference between a tool that can hold your whole problem in its head and one that keeps forgetting. If it delivers, it is a genuinely big deal.

The operative word being "if."

The two-tier rollout is the real story

I want to flag the pattern here, because it keeps happening across the industry. The frontier model lands for the biggest enterprise customers first, and everybody else waits. Vertex AI customers get hands-on access while the solo operator and the small shop get a blog post and a "coming soon."

I get the business logic. Enterprise customers pay the bills, and a controlled rollout is safer. But let’s be honest about what it means for the rest of us. The most capable AI is increasingly gated behind enterprise contracts at launch. The democratization story we were sold is real eventually — it is just not real on day one anymore.

What this means for you

If you are a small business or solo operator: do not build your workflow around a model you cannot actually get. I see people reorganizing their whole stack in anticipation of an unreleased model. Do not. The tools you can use today — Claude, GPT, the current Gemini — are extraordinary right now. Ship with what is in your hands. When 3.5 Pro actually opens up, you can evaluate it then, on real performance, not a keynote promise.

If you are in enterprise IT: this is your window, and it cuts both ways. Early Vertex access is a genuine advantage — pilot it, benchmark it against what you are running, and find out whether 2M context changes anything for your real workloads. But do not let "we have early access" turn into a production dependency on a model that is still officially in preview. Preview means it can change, get repriced, or wobble under you. Test hard. Commit slowly.

For everyone: take launch timelines as marketing, not as a calendar you can plan against. "Next month" became "still waiting" in the span of one month. That is not unique to Google — it is the whole industry. Plan around what ships, not what is promised.

My take

I have not tested Gemini 3.5 Pro, because I cannot — and that is the whole point of this piece. I am not knocking the model. By all accounts it will be excellent, and that 2M context window genuinely interests me. What I am knocking is the growing gap between the announcement and the availability. We are in an era where AI labs get enormous credit for unveiling things, and the actual shipping quietly slips.

So here is my rule, and I would offer it to you: a model is not real until you can put your own work into it. Until then, it is a press release. Keep your hands on the tools you actually have. There are plenty, and they are more than good enough to be transforming your business right now.

News commentary by Brad Rowland — IT Infrastructure and Operations leader, automation builder, and AI implementer. Sources are linked inline.