Gemini 3.5 Pro
Google announced Gemini 3.5 Pro at I/O on May 19 with a June GA target — then the date slipped, and so did the leaked July 17 launch attempt (codename “Cappuccino”) and its July 24 fallback. Google says it is “currently testing with partners,” and on Sep 24, per 9to5Google, DeepMind said Google had stepped back from it to focus on Flash. On Sep 30, the day Google announced Gemini 4 Argon, 9to5Google described 3.5 Pro as cancelled; Google itself has not said so. There is no general availability date. Whenever it lands, the second it answers on the public API, ArtificialWatch fires.
Leaks & news, newest first.
Reported cancelled as Google launches Gemini 4 Argon
9to5Google’s Argon launch story says the new model “follows the cancellation of Gemini 3.5 Pro.” Google’s own Argon post does not mention 3.5 Pro, and no Google statement on it has been published. Posts on X calling it “officially cancelled” trace back to fan accounts, not Google. Our hunt stays armed in case a 3.5 Pro id ever ships. via 9to5Google
Google steps back from 3.5 Pro to focus on Flash
Speaking at The Information’s AI Agenda summit, DeepMind chief Koray Kavukcuoglu said Google stepped back from Gemini 3.5 Pro to focus on its Flash models, per 9to5Google’s account, which notes Google gave no word on whether 3.5 Pro will still launch. Gemini 4 is now in post-training, with an early version planned “as soon as possible.” via 9to5Google
Google: 3.5 Pro is “testing with partners”
In the post launching Gemini 3.6 Flash, Google wrote that “Gemini 3.5 Pro is currently testing with partners and we plan to make it broadly available as soon as it’s ready” — no date. The same post said Google had started its “most ambitious pre-training run yet, for Gemini 4.” via Google
"Cappuccino" leak wave: July 17 target, 2M context, Deep Think
A fresh round of leaks says Gemini 3.5 Pro — internal codename Cappuccino — is targeting July 17 with July 24 as the fallback: an entirely new pretraining run, a 2M-token context window, a Deep Think reasoning mode, and big claimed jumps in web dev, UI generation and SVG work. The same leaks claim it beats GPT-5.6 and Claude Fable 5 in internal evals — vendor-internal numbers, unconfirmed by Google, so grade accordingly. Notably, the market disagrees on timing: July buckets have collapsed and the weight sits in August. via @pankajkumar_dev on X
Second week of July, still in preview
Gemini 3.5 Pro entered the second week of July in limited preview — no confirmed GA date, no published benchmarks, no final pricing. Enterprise teams are being told to evaluate on the preview at their own risk. via MarketScale
July 17 target — and a ground-up architectural rebuild
Reporting pegs the new launch window at July 17 and says Google abandoned the 2.5-era architecture for a ground-up redesign. Unconfirmed by Google — treat the date as sentiment, not schedule. via BigGo Finance
Why it slipped: tokens, coding, long-horizon reasoning
Early testers reportedly flagged token-efficiency concerns, coding performance below flagship bar, and long-horizon multi-step reasoning short of what I/O promised. Four senior Gemini researchers reportedly departed for Anthropic in the same stretch. via Bind AI
Unveiled at I/O with a June GA target
Google announced Gemini 3.5 Pro on the I/O mainstage and said it was targeting June for general availability — the target it has since walked past. via Google I/O
The official word.
"Gemini 3.5 Pro is currently testing with partners and we plan to make it broadly available as soon as it’s ready."
Google, launching Gemini 3.6 Flash · Jul 21, 2026 — no GA date given
Hear it the second 3.5 Pro goes GA.
We sweep Google's public API every 60 seconds. The moment Gemini 3.5 Pro answers outside the preview gate, your channels fire — email + Chrome push free, SMS and a phone call on paid tiers.