What the Rumors Say
Synthesizing recent days' reports from TechTimes, Geeky Gadgets, and others, Google DeepMind has internally set Gemini 3.5 Pro's launch target for July 17, slipping to July 24 if more time is needed. Rumored specs: a 2-million-token context window (the largest among mainstream flagships if true), a Deep Think deep-reasoning mode (said to be tied to the $250/month AI Ultra tier), and a base model retrained from scratch. Leaks about API pricing contradict each other—one says $12–15/M input, another says $1.25/M—which itself shows the sources' reliability is limited.
What Can Be Confirmed
Only three things can be pinned down: the model does exist and is running internally; the launch has slipped from June to July (Pichai hinted at a delay at Google I/O on May 19); and as of press time, the public Gemini API docs contain no model card or pricing page for gemini-3.5-pro. Everything else—including the claim that "internal evaluations lead GPT-5.6 Sol and Claude Fable 5"—is unconfirmed.
Crashing Into a Crowded Window
If the 17th is real, it will squeeze into an unusually crowded window: GPT-5.6, Grok 4.5, and Muse Spark 1.1 all launched in the past ten days, and the same day is also the opening of Shanghai's WAIC. The pragmatic advice for developers: don't make technology-selection or cost calculations based on leaked specs; wait for the official model card and pricing page before acting—this time even the date could slip again.
via: TechTimes (July 13), Geeky Gadgets, and other media leak roundups—none confirmed by Google officially