Google has pushed the general availability of Gemini 3.5 Pro from June into July 2026, according to reporting from Business Insider confirmed by multiple outlets. The model remains in limited preview for select Vertex enterprise customers while Google collects tester feedback and makes adjustments.

The delay is modest in calendar terms — a few weeks — but its timing matters. Google CEO Sundar Pichai announced at I/O in May that Gemini 3.5 Pro would arrive “next month.” Enterprise teams who built deployment plans around a June go-live now need to extend their timelines.

What Gemini 3.5 Pro Was Supposed to Deliver

The headline specs have not changed. Gemini 3.5 Pro is still targeting a 2 million-token context window and a Deep Think reasoning mode, placing it in direct competition with the top tiers of Claude and GPT.

The 2 million-token context window is the capability enterprise buyers are most interested in. At that scale, a model can ingest an entire company’s contract library, a full year of support tickets, or a large codebase in a single call. That is not a marginal improvement over current context limits — it changes the category of problem you can solve.

Deep Think, Google’s name for extended reasoning, is being gated behind the Ultra subscription tier at $250 per month — making it the most expensive consumer AI subscription in the market by a meaningful margin. For enterprise teams evaluating the full pricing picture, Deep Think access will require a separate budget line.

On standard pricing, Gemini 3.5 Pro is expected to come in at $15 per million input tokens and $60 per million output tokens — positioning it as a premium model aimed at high-stakes reasoning tasks, not everyday throughput.

Why the Delay Happened

Google is using the additional window to collect structured input from early testers and make targeted refinements to the model. That is not unusual for a frontier release — Anthropic did the same before Fable 5 went public this month.

What is unusual is the context around the delay. Google’s AI team has seen notable researcher departures in recent months, and several are reported to have joined Anthropic and OpenAI. The company is also navigating the internal complexity of having multiple flagship AI product lines — Gemini in Search, Gemini for Workspace, Vertex AI for enterprise — that all need to coherently support a new flagship model at launch.

None of that means Gemini 3.5 Pro will be a weaker model. It means Google is taking launch quality seriously at a moment when it cannot afford a rough debut.

What This Means for Business

If you were waiting for Gemini 3.5 Pro before committing: The decision just got a month simpler. The July window is not indefinite — it is a defined, short delay, not a shelving.

If you are already evaluating Claude Fable 5 or GPT-5 equivalents: The delay does not change the competitive landscape dramatically. By July you will have three comparable frontier models in market, all with long-context reasoning. The differentiator will be pricing, tooling, and how well each integrates with your existing stack.

If you are on Google Workspace or Vertex already: The smart move is to request access to the limited preview now if you have not already. Early testers are shaping the final model — being in that group gives you both influence and a head start.

For enterprise data teams: The 2 million-token context window is the more important number than any benchmark result. If your use case involves reasoning across large document sets — contracts, financial filings, customer histories, codebase — Gemini 3.5 Pro becomes relevant when it ships, regardless of which week that is.

The Bigger Picture

Frontier AI model releases are becoming increasingly complex events. The gap between “announced at conference” and “available in production” is growing, not shrinking, as models become more capable and the stakes of a bad launch increase.

Anthropic spent weeks rolling out Fable 5 access and has been adjusting pricing as it goes. Google is choosing to get the model right before opening it up. OpenAI has been similarly cautious with its most capable releases.

For enterprise buyers, this pattern has a practical implication: stop building deployment timelines around conference announcements. Build them around GA dates — and expect those dates to slide by a few weeks even after they are announced.

The frontier model market is delivering genuinely remarkable capabilities. It is also delivering them on a less predictable schedule than the launch keynotes suggest. Planning for that reality is now part of enterprise AI strategy.

---

If you’re deciding where to start with agents, start here. The free Working With Claude field guide walks through the ecosystem, Claude Code, and a real rollout plan. Get your copy.