After three missed deadlines, a scrapped base model, and months of quiet frustration, the Gemini 3.5 Pro saga appears to have reached its final chapter — and it’s not the ending anyone expected. Rather than finally shipping, reports now suggest Google has simply cancelled the model altogether, skipping straight to its next major project instead. Here’s what actually happened, and why this decision landed alongside one of the biggest leadership shake-ups in Google’s AI history.
From “Coming Next Month” to Cancelled Entirely
The story of Gemini 3.5 Pro started with confidence, not caution. At Google I/O in May 2026, Sundar Pichai told a room full of developers that Gemini 3.5 Pro was coming, asking the crowd for patience and promising it would arrive “next month” — a line that reportedly drew an audible groan from the audience. Google shipped Gemini 3.5 Flash that same day instead, a lighter model built for agentic and coding tasks, while the flagship Pro model stayed just out of reach.
June came and went with no Pro model. Google spent July retraining Gemini on updated data specifically to fix coding performance, and when that work fell short of internal expectations, the company shipped Gemini 3.6 Flash on July 21 as a stopgap. A widely tracked July 17 target came and went too, with the rebuilt model reportedly still plagued by hallucinations and inconsistent outputs that prevented it from clearing basic reliability standards.
Now, according to a new institutional note from Semi Analysis, Google has quietly shelved Gemini 3.5 Pro entirely and is pivoting to hype around Gemini 4 instead. As of publication, the Gemini API release notes still carry no entry for a Gemini 3.5 Pro model at all — a detail that lines up with the cancellation theory rather than an unusually long delay.
Why the Model Reportedly Got Scrapped, Not Just Delayed
The most important detail in this story isn’t just that the model is gone — it’s why. Semi Analysis argues that whatever structural issues held back Gemini 3.5 Pro, particularly around coding and agentic reliability, aren’t the kind of problems a bigger pre-training run automatically resolves. In plainer terms: the flaws weren’t things Google could simply patch by throwing more compute at the same underlying architecture.
That framing matters because it changes how much confidence the industry should place in Gemini 4 solving the same issues. Google has been open about lagging behind on agentic coding specifically, going as far as forming a dedicated DeepMind coding team to close the gap — an acknowledgment that the weakness runs deeper than one unlucky model generation.
Separately, reporting from Geeky Gadgets pointed to a mix of development delays and internal obstacles behind the decision, suggesting Google is reallocating resources directly toward Gemini 4 rather than continuing to sink further effort into fixing 3.5 Pro. The move is also being read as a response to mounting competition from rivals, whose models have reportedly outperformed the Gemini series in key recent benchmarks.
The Leadership Shake-Up That Landed at the Same Time
This cancellation didn’t happen in isolation — it arrived in the middle of one of the most significant leadership changes at Google DeepMind in years. Demis Hassabis, DeepMind’s co-founder and CEO since Google absorbed the company in 2014, is stepping back from day-to-day operational control, transitioning to become Chair of Google DeepMind and taking on a newly created role as Chief Scientist of Alphabet.
Taking over the reins is Koray Kavukcuoglu, DeepMind’s CTO for the past thirteen years, who is stepping up as Senior Vice President to lead AI research and Gemini model development directly. In company briefings, Kavukcuoglu had reportedly already been leading Gemini discussions in practice, with Hassabis attending only the most significant meetings — suggesting this formal handover reflects a shift that was already underway internally.
The reshuffle didn’t stop there. Jeff Dean, Google’s chief scientist and a 27-year veteran of the company, is also leaving — departing to co-found Discovery Loop, a public benefit corporation aimed at automating machine learning, science, and engineering research, alongside fellow former Googlers Sanjay Ghemawat, Oriol Vinyals, and Quoc Le. Notably, Alphabet itself is a founding investor in Dean’s new venture and will supply it with cloud compute, a detail that softens what might otherwise look like a purely adversarial departure.
Hassabis, for his part, framed the transition around a bigger long-term focus rather than retreat. In his note to staff, he cited the unreleased Gemini 4 by name, expressing confidence in the “great progress” being made, and said he plans to devote more time to Isomorphic Labs, Alphabet’s AI drug discovery unit, which he has described as one of the most important applications of AI for improving human life.
Markets didn’t take the news calmly. Alphabet shares fell around 5% following the announcement of the reshuffle, reflecting real investor concern about losing two of the company’s most senior AI leaders at exactly the moment Google is trying to close a competitive gap.
What This Means for Gemini 4
With Gemini 3.5 Pro reportedly off the table, all eyes are now on Gemini 4 — and the pressure on it to deliver is enormous. Google has confirmed it has started training Gemini 4, positioning it as the next major frontier model expected to redefine the company’s AI capabilities and help it reclaim a competitive position it’s visibly lost ground on over the past year.
But skepticism is warranted here too. The Semi Analysis report isn’t optimistic that Gemini 4 automatically fixes the underlying issues, since the same structural weaknesses in coding and agentic reliability that sank 3.5 Pro aren’t the kind of problems that scale away with a bigger model. Until a released version proves otherwise, the report reportedly ranks Gemini in eighth place overall among current frontier models — a strikingly low position for a company with Google’s resources and research history.
A Pattern of Struggling to Ship, Not Struggling to Build
It’s worth stepping back and noting what this saga is really about. Google hasn’t struggled to produce impressive AI research — DeepMind’s broader track record, including Hassabis’s own Nobel Prize-winning work on protein structure prediction, proves the underlying talent and capability exist. What this cancellation highlights instead is a harder, more operational problem: turning ambitious research into a shipped, reliable flagship product on a competitive timeline, especially as rivals continue releasing capable models at a faster pace.
Conclusion
Gemini 3.5 Pro’s story didn’t end with a late but triumphant launch — it ended with a cancellation, a leadership shake-up, and a very public bet that Gemini 4 can succeed where its predecessor couldn’t. Whether that bet pays off depends on whether Google’s new leadership under Koray Kavukcuoglu can solve the coding and agentic reliability issues that reportedly doomed the last model, or whether Gemini 4 simply inherits the same problems at a larger scale. For now, this stands as one of the most dramatic model cancellations the AI industry has seen, and a clear signal that even a company with Google’s resources isn’t immune to shipping failures under competitive pressure.
Read More :- AI Agents Are Becoming a Bigger Deal — Here’s What That Actually Means | Affitronix
FAQs
Q1: Has Google officially confirmed that Gemini 3.5 Pro was cancelled?
No. As of the latest reporting, Google has not issued an official statement confirming the cancellation. The information comes from an institutional note by Semi Analysis and other reports citing people familiar with the situation, and the Gemini API release notes still show no entry for a 3.5 Pro model.
Q2: Why was Gemini 3.5 Pro reportedly cancelled instead of just delayed further?
Reports suggest the model suffered from structural issues in coding and agentic reliability that couldn’t be fixed simply by retraining or scaling up, leading Google to redirect resources toward Gemini 4 instead of continuing to patch the existing model.
Q3: Who is taking over Gemini development after Demis Hassabis stepped back?
Koray Kavukcuoglu, DeepMind’s CTO for the past thirteen years, is taking over as Senior Vice President to lead AI research and Gemini model development, while Hassabis becomes Chair of DeepMind and Chief Scientist of Alphabet.
Q4: When is Gemini 4 expected to be released?
Google has confirmed it has started training Gemini 4, but no official release date has been announced. Analysts note that its success is seen as critical for Google to regain competitive ground in the AI industry.
