Google announced Gemini 3.5 Pro at I/O on May 19 and committed to general availability in June. As of June 21, the model is still gated behind a limited preview for selected Vertex AI enterprise customers and has not shipped to the public Gemini app, Google AI Studio, the consumer Gemini subscription, or the general developer API. Google has not issued a public update on the delay. The Flash variant, by contrast, has been generally available since shortly after I/O.
Pro is the model on which Google's competitive narrative for the second half of the year depends. The announced specifications include a two-million-token context window, a 'Deep Think' extended-reasoning mode, and upgraded multimodal handling. None of those features are testable outside the enterprise preview, which has so far surfaced limited third-party benchmarking and no apples-to-apples comparison against Claude Fable 5 or OpenAI's GPT-5.5 Instant on coding, agentic-task, or long-context evaluations.
The slip lands in an unusually crowded month. Anthropic's Fable 5 went broadly available in Copilot tiers on June 9 and reached 83.1% on Terminal-Bench 2.1. OpenAI shipped GPT-5.5 Instant as the new default ChatGPT model. Microsoft, NVIDIA, and a handful of Chinese labs have all moved capable models into the same window. Each week Pro stays in preview is a week competitors get to define the frontier story without it.
Takeaway for learners: announcement dates and shipping dates are not the same thing, and the gap between them is one of the more reliable signals in AI. When a lab demos a capability but cannot put it behind a public API within the promised window, the most common explanations are infrastructure capacity, safety review, or a benchmark that did not survive contact with a broader population of users. If you are deciding what to build on, pay more attention to what is actually callable today than to what was on the slide last month.