Status as of 24 June 2026: Gemini 3.5 Pro was announced at Google I/O on 19 May 2026 with a June GA target. It is available only to a small number of enterprise customers on Vertex AI preview. It is not in the consumer Gemini app, not in Google AI Studio's public model picker, and not in the general API. Six days remain in June.
What Gemini 3.5 Pro is supposed to be
The specification announced at I/O positions Gemini 3.5 Pro as Google's frontier model for complex, extended business tasks. The confirmed or widely reported features include:
- 2-million-token context window — the largest of any commercially available model. Capable of holding approximately 1,500 pages of text, a full year's worth of business emails, or a substantial codebase in a single context.
- Deep Think reasoning mode — an extended inference mode for complex multi-step tasks, similar to OpenAI's o1 and Anthropic's extended thinking.
- Pricing announced at approximately £2.50 per million input tokens / £15 per million output tokens — competitive with Claude Sonnet 4.6 for extended document tasks.
- Google Workspace integration — direct integration with Docs, Gmail, Sheets, and Meet for business users on Workspace Business and Enterprise plans.
The model is positioned between Gemini 3.5 Flash (the fast, cheap, already-GA model) and a potential Gemini 4 release later in 2026.
Why it is still in limited preview
The delay is likely due to benchmark performance not yet clearing Google's internal release bar for a model carrying the headline "Pro" label. Gemini 3.5 Flash — the lighter model — launched at I/O and has been the default in the Gemini app since May. The Pro version carries higher expectations and, given the competitive pressure from GPT-5.6 Kindle (released or imminent) and Claude Fable 5, a disappointing launch would be more damaging than a quiet delay.
Prediction markets on Polymarket place the probability of a June 30 GA release at 50 to 55 percent. Several AI analysts tracking Vertex AI preview access report "strong performance" on extended document reasoning and coding tasks, which suggests the quality bar is clearing — the delay is more likely release process and infrastructure scaling than fundamental model problems.
What UK operators should actually do
Gemini 3.5 Pro matters most to you because of the Workspace integration. When it goes GA, you will be able to use it directly in Docs, Gmail, and Sheets without a separate API subscription. If you are on Workspace Business Standard or higher, watch the official Google Workspace blog for the GA announcement. Do not extend or renegotiate your Workspace plan based on the Pro features until the GA launch is confirmed — the spec could change.
Do not wait for Gemini 3.5 Pro before making a decision. The model is not currently available, and the six-day window is genuinely uncertain. Make your current workflow decision based on what is available now — Claude Sonnet 4.6, GPT-5.5 or GPT-5.6 (if released), Gemini 3.5 Flash — and plan a reassessment in mid-July once the Gemini 3.5 Pro situation has resolved one way or another. A workflow decision delayed waiting for a model release is lost productivity every week it waits.
Check the model picker in AI Studio daily this week. When Gemini 3.5 Pro goes GA, it will appear there first. Run your standard test task on it — the same task you use to benchmark other models — and compare against your current tool. The 2-million-token context is genuinely useful for service businesses that process long documents (contracts, survey results, email threads, inspection reports). If that is your use case, the Pro model is worth evaluating seriously.
The broader context
The Gemini 3.5 Pro delay illustrates a pattern that is becoming common in 2026: model announcements are made at developer events weeks or months before GA availability, and the gap between "announced" and "available" is creating decision uncertainty for businesses trying to plan their AI tooling.
The practical response is to separate two decisions: the "start now" decision (which model you use for your first or current AI workflows, based on what is available today) and the "platform commitment" decision (which vendor you standardise on, which should wait for a clearer picture of the full model landscape). Make the first decision immediately. Defer the second until July, when both the Gemini 3.5 Pro situation and the GPT-5.6 Kindle release will have resolved.
What to do
Operator action
Today: Set a calendar alert for 1 July. If Gemini 3.5 Pro has launched, run your test task on it. If not, note it for your July AI tool review.
This week: Do not delay current AI workflow decisions waiting for Gemini 3.5 Pro. Use what is available now.
Google Workspace users: Watch the Google Workspace blog and AI Studio model picker this week. The launch, if it happens, will appear there first.
Platform decisions: Do not extend multi-year Google AI or workspace contracts this week based on anticipated Pro features. Wait for GA.
The model is likely very good. It may arrive this week or it may arrive in July. Either way, the businesses that start AI workflows now on available tools will be ahead of the ones waiting for the perfect model to be released.