Google is on the verge of shipping Gemini 3.5 Pro, the flagship model it unveiled at its I/O developer conference on May 19 and promised for general availability in June. Chief executive Sundar Pichai asked the I/O audience to wait roughly another month for access, a delay that reportedly drew audible groans from a crowd hoping to use it immediately.
What the model brings
According to reporting from TechTimes and others, Pro targets a 2 million token context window, large enough to hold entire codebases, lengthy contracts or whole document sets in a single prompt, alongside a reasoning mode Google calls Deep Think and frontier grade multimodal understanding across text, images and other formats. Pricing is expected to land at roughly ten times that of the already released Gemini 3.5 Flash, with estimates near 15 dollars per million input tokens and 60 dollars per million output tokens, and early access arriving first through Google's 20 dollar Pro and 250 dollar Ultra consumer tiers.
The race it belongs to
The launch slots into an intensifying contest among the frontier labs. Flash went generally available at I/O at 1.50 and 9.00 dollars per million input and output tokens and roughly four times the speed of the prior generation, while OpenAI and Anthropic press on with their own agent platforms and Microsoft and Google have pushed into AI coding models that the two younger labs largely defined. Elon Musk's xAI, by contrast, is still training Grok 5, with no confirmed release before the end of June.
Why scale matters for work
For the labour market, the significant feature is not raw benchmark scores but the combination of a vast memory and stronger step by step reasoning. A model that can ingest an entire body of material and reason across it in one pass is a far closer substitute for the research, synthesis and first draft analysis that juniors and analysts have traditionally done than a short context chatbot ever was.
A caveat worth keeping
One important note tempers the hype: as of early June the specifications remained vendor claims rather than independently tested results, since Pro had not yet shipped to the public. The pattern across the AI economy holds regardless. Each leap in capability tends to widen the set of tasks employers feel comfortable handing to software, and to sharpen the premium on the people who can direct, check and govern what these systems produce.
