Video: "Gemini 3.5 Pro DELAYED Again... BUT Gemini 3.6 Flash Might Drop Soon!" by Julian Goldie on YouTube.
Three missed deadlines
Gemini 3.5 Pro appeared briefly as a limited preview in May 2026 but was not given a full release. A June general availability date came and went without a launch. The 17 July target followed — also missed. Bloomberg and CNBC both reported the third slip on 16 July 2026.
Google has not made a public statement on a new timeline. The absence of any replacement date is notable: it suggests the team does not yet have a firm completion point rather than simply choosing not to communicate one.
Why it keeps slipping: coding performance
The stated reason for the hold is that Gemini 3.5 Pro does not yet consistently hit the coding benchmark scores Google requires before putting it in front of users. Coding is one of the primary use cases Google is positioning the model for, which makes it a hard problem to ship around.
That is the internal framing, at least. External observers have noted that the gap between the leaked capabilities — 2-million-token context, Deep Think reasoning mode, the Spark agent — and what a production model reliably delivers is often wider than internal benchmarks suggest, particularly for coding tasks at scale.
Talent leaving DeepMind
Four senior DeepMind researchers have departed around the same period. The most prominent are Noam Shazeer, who went to OpenAI, and John Jumper, who joined Anthropic. Both are significant research contributors. The departures do not explain the delay on their own — both left after the model had been in development for some time — but they have added to investor concern about Google's AI team stability.
Alphabet shares fell roughly 5% on the news of the delay, which represents approximately $200 billion in market capitalisation. That reaction reflects how much of Google's near-term search advertising story depends on Gemini 3.5 Pro actually shipping.
Gemini 3.6 Flash as a possible interim release
Rather than wait for Gemini 3.5 Pro, Google is reportedly considering releasing Gemini 3.6 Flash first. Flash models sit below Pro in the Gemini hierarchy — faster responses, lower cost, lower capability ceiling. The idea is to maintain visible forward momentum while the Pro version continues development.
Gemini 3.6 Flash would likely become the model powering AI Overviews and AI Mode in Google Search in the interim. That matters for any business whose content currently appears in AI-generated search summaries: the model producing those summaries would be different from the one that was originally anticipated, with a different context window and different reasoning behaviour.
What this means for businesses using Google and AI search
Businesses that were planning for Gemini 3.5 Pro features — particularly the Spark agent's ability to pull from Gmail, Docs and Drive, or the Deep Think reasoning mode — will need to treat those as further away than previously expected. There is no indication of when Pro will actually ship.
AI Overviews in Google Search are not going away in the meantime; they continue to run on existing Gemini models. But if 3.6 Flash becomes the next deployed model, the characteristics of AI search results — what gets cited, how long those citations are, how the model handles ambiguous queries — will shift in a direction that is not yet fully documented.
The practical response is not to optimise for a specific model version. It is to maintain content that is factually accurate, structured clearly, and directly answers specific questions — the kind of content that tends to appear in AI-generated summaries regardless of which model is producing them.
Where this connects to NordSys
NordSys helps UK businesses stay visible as Google's AI search stack keeps moving. The delay to Gemini 3.5 Pro is one more reason why ranking strategies built around a fixed set of model assumptions do not hold. We build SEO and content approaches that work with current AI Overviews and adapt as the underlying models change — without requiring you to track every announcement yourself.
See our SEO & AI Ranking service →