Reports point to July 17 for general availability, but Google has never confirmed the date.
As of July 13 the model was not listed in the public Gemini API at all.
The widely quoted 2 million token context window appears in no official Google documentation.
Half the internet says Google's biggest model of the year ships today. Google has not said it once.
What is actually confirmed
Google announced the Gemini 3.5 family at I/O on May 19 and shipped only Gemini 3.5 Flash that day. Pichai told developers to give Google until next month for Pro, a line that reportedly drew an audible groan from the room.
June came and went. Google pushed general availability to July, citing quality refinements after early enterprise testing. Reports were blunter, flagging that Pro was struggling on advanced reasoning, coding and long-horizon task execution, with token-efficiency problems.
As of July 13, Gemini 3.5 Pro was not listed as a generally available model in the public Gemini API. Google AI Studio and the API showed gemini-3.5-flash and gemini-3.1-pro-preview. A limited enterprise preview has run on Vertex AI since at least late June.
Third-party reporting targets July 17. Google has not confirmed this.
Treat every spec as a rumour
This is where most coverage fails you. The 2 million token figure is widely reported but not confirmed by Google in any official documentation. Same for the $250 per month Ultra tier gating Deep Think, and the roughly $1.25 input and $10 output per million token pricing.
One analyst put the caution well: until the model is generally available and independent testers can run their own evaluations, its advertised capabilities remain Google's claims rather than verified results. Advertised context window and effective context performance are also different things, and a 2M window that degrades badly at 800K is not a 2M window in practice.
The July 17 date may also refer to a preview, a limited rollout, an LMArena appearance, or a staggered release rather than true general availability.
Why it matters anyway: Pro is six weeks late into the most competitive week AI has had, landing after GPT-5.6 and Grok 4.5. Google has been losing people too, with Noam Shazeer going to OpenAI and Nobel laureate John Jumper leaving for Anthropic. A great model quiets all of it. Another slip does not.
Quick Links:
