Gemini 3.5 Pro's 'next month' has arrived — and Pro hasn't

At Google I/O on 19 May, Google did two things with Gemini 3.5. It shipped Flash — fast, cheap, and immediately the default in the Gemini app and AI Mode in Search. And it announced Pro — the larger, more capable sibling — without releasing it, saying the model was already 'being used internally' and would roll out 'next month'. Five weeks later, the cheap one is everywhere and the powerful one is still a promise. 'Next month' has become this month, and Pro is not here.

It is worth being precise about the promise, because the gap is between Google's own words and Google's own calendar — not between a rumour and reality. Google's I/O materials announced 3.5 Pro, described it as in internal use, and gave a 'next month' rollout framing. That is a soft commitment, not a hard date, and soft commitments slip. But a 'next month' said in mid-May points squarely at June, and June is nearly gone. The honest status is therefore narrow and defensible: Pro was announced, Pro is reportedly in internal use, and Pro has not shipped to the public on the timeline Google itself set.

Much of the coverage attaches concrete numbers to 3.5 Pro: a context window around 2 million tokens and a 'Deep Think' reasoning mode. These are worth flagging carefully, because they are expectations in the trade press, not figures Google has confirmed for 3.5 Pro. The distinction matters for anyone making a build decision. A 2M-token context is a meaningful capability — roughly double what most frontier models expose — but quoting it as a settled fact today over-states what is known. The same caution applies to 'Deep Think': it is a plausible direction given Google's reasoning work, not a confirmed feature of an unreleased model.

A delayed frontier model is not a scandal. The larger a model, the longer the tail of evaluation, safety testing and serving work between 'trained' and 'available', and Pro-tier models routinely trail their Flash-tier siblings. What the slip does tell you is about sequencing and confidence: Google chose to ship the cheap, fast model first and hold the flagship, which is a reasonable way to get a capable model into the most hands quickly while the heavier one is hardened. It also tells you that the 'next month' framing was optimistic — a useful calibration the next time a launch window is floated from a keynote stage.

The actionable read: if your plan depends on 3.5 Pro specifically, it is not yet a dependency you can take — evaluate Flash, which is real and strong for its tier, and treat Pro as a watch item. The moment to form a view on Pro's quality is when a stable endpoint and a model card with benchmark numbers appear. Until then, the most accurate thing that can be said about Gemini 3.5 Pro is that Google says it exists, Google says it is nearly here, and Google said that last month too.