Google is reportedly targeting July 17 for the launch of Gemini 3.5 Pro — but treat that date, and everything else in this piece, as a leak, not a confirmed roadmap. As of July 13, there’s no model card, no pricing page, and no gemini-3.5-pro listing anywhere in Google’s public API docs. That’s the honest starting point for the Gemini 3.5 Pro launch date rumor that’s been spreading across leak-tracking sites and X all week.
Why care about a date nobody at Google DeepMind has confirmed? Because if the leaks are even directionally right, Gemini 3.5 Pro lands with a meaningfully bigger context window and a brand-new reasoning mode right as two rival flagship models are already on shelves. Google would be arriving late to a race it used to set the pace for.
What’s Reportedly Coming Before the Gemini 3.5 Pro Launch Date
The specs circulating are consistent enough across independent leak sources — Geeky Gadgets, AIToolsRecap, and a handful of well-followed X accounts — that the convergence itself is part of the story. Multiple unrelated leakers landing on the same numbers is usually a decent signal, even without an official source. The headline figure is a 2 million token context window, roughly double what most flagship models ship with today — enough to hold an entire mid-size codebase or a stack of research papers in memory at once, without the usual chunking workarounds developers build around smaller windows.
The second piece is a mode leakers are calling “Deep Think” — a slower, more deliberate reasoning layer for multi-step problems, in the same spirit as the extended-thinking modes other labs rolled out this year. Here’s the catch: it’s genuinely unclear whether Deep Think ships on day one or arrives later as a separate rollout. None of the sourcing nails that down, so take it with a grain of salt until Google says otherwise.
There’s also a wrinkle that helps explain why this is cutting it close. Google reportedly scrapped the original Gemini 3.5 Pro base model partway through training and restarted pretraining from a cleaner setup. If that’s accurate, it’s exactly the kind of thing that pushes a launch date around at the last minute — and it’s a big part of why some coverage treats July 17 as near-certain while other outlets keep flagging it as unconfirmed until Google actually publishes documentation.
How the Rumored Gemini 3.5 Pro Launch Date Fits the July AI Race
Context matters here, because July has already been a packed month for frontier models. OpenAI has had its Sol/Terra/Luna generation — what most people just call GPT-5.6 — in preview since June 26; we covered everything OpenAI shipped with GPT-5.6 if you want the full breakdown of what changed. xAI followed with a general release of Grok 4.5 on July 8, a launch we put through its paces in our review of xAI’s Grok 4.5. If you’re trying to figure out how any of this stacks up for real coding work rather than benchmark charts, how Claude Sonnet 5 stacks up for developers is worth a read too — it’s the model most leaked Gemini 3.5 Pro numbers are quietly being measured against.
Pricing is the other rumor worth flagging honestly as a rumor, because the numbers don’t even agree with each other yet. One leak points to something like $12-15 per million input tokens and $36-45 per million output tokens — noticeably cheaper than the roughly $15/$60 split other estimates are floating for the same model. Two competing guesses that far apart is usually a sign nobody outside Google actually knows the real number yet. It’s worth revisiting once — if — official pricing shows up, since it’ll shape whether Gemini 3.5 Pro undercuts GPT-5.6 and Grok 4.5 or lands right in line with them.
What’s Next
The honest read: every claim in this piece, from the July 17 date down to Deep Think’s launch-day status, should be treated as unconfirmed until Google publishes an actual model card. The pretraining restart alone is reason enough to expect the date could still slip a week or two, the way big model launches often do when a lab restarts training mid-cycle. Google isn’t the only lab racing a self-imposed July deadline, either — DeepSeek reportedly has its own target of July 24 for a competing release, which tells you roughly how compressed this month has gotten for everyone building frontier models right now.
For comparison, the currently shipped Gemini 3.1 Pro is still what most benchmark pieces are testing against today, which is a reminder that everything about “3.5 Pro” so far is genuinely forward-looking rather than confirmed reality. If Google does hit July 17, expect the specs, real pricing, and a head-to-head against GPT-5.6 and Grok 4.5 the same day. Until then, this is a leak worth watching closely, not a launch worth planning your workflow around just yet.
Deixe um comentário