Gemini 3.5 Pro: Every Leak Before Thursday’s Expected Launch

Gemini 3.5 Pro is expected to launch on Thursday, July 17, 2026 — two days from now, though Google has not confirmed the date. Leaked reports point to a 2-million-token context window, a Deep Think reasoning layer reserved for the $250-per-month Ultra tier, and a model rebuilt entirely from scratch — not an incremental upgrade from Gemini 2.5 Pro — after running more than six weeks behind its original late-June target.

What the Gemini 3.5 Pro Leaks Actually Say

As reported by TechTimes on July 13, Gemini 3.5 Pro is targeting a July 17 release after a full architectural rebuild — and every single spec remains unconfirmed. The headline leak is a 2-million-token context window, which would be the largest of any flagship model. For comparison, Anthropic’s Claude Fable 5 ships with a 200K standard window, and OpenAI has not publicly confirmed the context size for GPT-5.6 Sol.

The second major leak concerns Deep Think, a reasoning layer that reportedly lets the model work through multi-step problems autonomously. Multiple reports place Deep Think exclusively on the $250-per-month Ultra plan — not the $20 standard subscription. If accurate, this continues the industry trend of frontier reasoning being priced out of reach for individual developers.

Reports also suggest the rebuild specifically targets three weaknesses of the previous generation: mathematical reasoning, SVG scene generation, and image quality.

On API pricing, a leaked post circulating on X suggests ~$12–15 per million input tokens and ~$36–45 per million output tokens. These figures are unconfirmed — treat them as developer speculation until Google publishes an official pricing page. For context, Gemini 3.1 Pro currently costs $2/$12 per million tokens; if the leak is accurate, Gemini 3.5 Pro would sit in a significantly higher tier, reflecting the compute cost of Deep Think reasoning.

Why Google Rebuilt the Model From Scratch

Gemini 3.5 Pro was originally slated for late June 2026 — making this launch more than six weeks late. According to StartupFortune, Google delayed the launch after scrapping its base model entirely, citing “quality refinements after early enterprise testing.” A full rebuild this late in a release cycle is unusual — it signals either serious performance problems in the original model or unusually high ambitions for the replacement.

Notably, as of July 15 there is no model card, no pricing page, and no gemini-3.5-pro listing in the public Gemini API documentation. Whatever ships Thursday, Google is keeping it sealed until the moment of launch.

The Talent Losses That Raised the Stakes

The delay arrived alongside the most significant talent exodus Google has seen in its AI division. On June 18, Noam Shazeer — Gemini co-lead and co-author of the original Transformer architecture paper — left for OpenAI. The following day, John Jumper, the 2024 Nobel laureate who led AlphaFold, announced his move to Anthropic, along with fellow researchers Jonas Adler and Alexander Pritzel.

The exits triggered a roughly 7% drop in Alphabet’s stock on June 22, erasing an estimated $270 billion in market value across two trading sessions — a stark signal of how tightly investor confidence in Google is now linked to its AI leadership pipeline.

The departures matter beyond the stock chart. Shazeer’s institutional knowledge of transformer architecture runs deep, and Jumper’s move to Anthropic directly strengthens the team building Claude. That both men left during the very period when Gemini 3.5 Pro was being rebuilt from scratch adds context the spec sheets won’t include. Thursday’s launch is not just a product release; it’s Google’s first major public signal that the rebuild was worth the cost of losing them.

Thursday Could Be the Biggest 48 Hours in AI This Year

July 17 is also opening day of the Shanghai 2026 World AI Conference, where Chinese President Xi Jinping is expected to attend in person for the first time — a direct signal that China treats AI as a national priority. Several Chinese model launches from Baidu, Alibaba, and others are expected at the event. If Gemini 3.5 Pro lands the same day, the two announcements together would make it the densest 48-hour stretch of AI news in 2026.

The timing follows a crowded frontier-model summer: OpenAI’s GPT-5.6 Sol, Terra, and Luna family entered limited preview on June 26, and Google’s own Gemini 3.5 Flash has been serving as the family’s speed tier since earlier this year.

What We Still Don’t Know

Almost everything. Google has confirmed neither the date, the pricing, the context window, nor the existence of Deep Think. The leaked API pricing of $12–15/$36–45 per million tokens — if real — would place Gemini 3.5 Pro at roughly 6–8× the cost of Gemini 3.5 Flash, which may reflect the reasoning overhead of Deep Think. There are no verified quotes from Google executives about the launch. If July 17 passes without a release, it would be at least the second delay — we will update this article immediately either way.

FAQ

Updated: July 15, 2026