Gemini 3.5 Pro tipped for 17 July launch with 2M-token window

A missed June deadline and a new July target

Google's Gemini 3.5 Pro — the company's flagship large language model and the most capable in the current Gemini generation — is widely reported to be targeting a general availability launch on 17 July 2026. That date has been cited by multiple third-party outlets including Business Insider and Geeky Gadgets, though Google has not officially confirmed it with a model card, API documentation, or pricing announcement. As of Sunday 12 July, the public Gemini API still lists only gemini-3.5-flash and gemini-3.1-pro-preview — not a generally available Pro 3.5 model.

The backstory involves a missed promise. At Google I/O on 19 May 2026, chief executive Sundar Pichai told the audience to

"give us until next month"

— a clear signal that Gemini 3.5 Pro would reach general availability in June. It did not. The June window closed with no public launch, and Google subsequently pushed availability to July, citing quality refinements following early enterprise testing.

Why it slipped — and what Google reportedly changed

The delay was not a simple polish job, according to multiple sources citing insiders. Rather than fine-tuning the previous Gemini 2.5 Pro as a base, Google's DeepMind team reportedly made the decision to abandon that iteration entirely and invest in a completely new pre-training process — a more expensive and time-consuming route, but one intended to produce a qualitative performance leap rather than incremental gains.

The reason, sources say, was significant performance ceilings in multi-step mathematical reasoning and SVG scene generation. Early enterprise testers reportedly found Pro trailing rivals including Claude Fable 5 and GPT-5.6 on coding and long-horizon reasoning tasks. None of this has been officially confirmed by Google.

What Google has actually confirmed

Two features were announced at I/O and are confirmed. First, a 2 million-token context window — double the 1 million tokens offered by Gemini 3.5 Flash and, according to Google's own framing, the largest of any production frontier model currently available. To put that in practical terms, 2 million tokens can hold roughly 1,500 pages of code, several full novels, or an entire enterprise codebase alongside its documentation. Second, a built-in Deep Think reasoning mode — Google's extended-reasoning capability, comparable in concept to Anthropic's extended thinking or OpenAI's high-effort reasoning, designed for the hardest abstract and mathematical problems.

Beyond those two confirmed features, reported specifications include autonomous workflow capabilities — letting the model manage coding tasks, tool usage, and execution with minimal human intervention — along with improvements in UI design generation and SVG vector graphic construction. These come from leaked test data and have not been officially verified.

Pricing: what's confirmed, what isn't

Google has published no official pricing for Gemini 3.5 Pro. Analyst sources project API costs significantly higher than the existing Gemini 2.5 Pro rates — roughly $15 per million input tokens and $60 per million output tokens, or in approximate UK terms around £12 and £48 respectively. These are analyst estimates only. By comparison, Gemini 3.5 Flash launched on 19 May at $1.50 input and $9.00 output per million tokens — confirmed figures.

On consumer subscriptions, Google made confirmed changes at I/O 2026. The Google AI Ultra plan was restructured: the entry tier dropped from $249.99 to $99.99 per month, offering roughly five times the usage limits of the standard Pro plan. A higher $199.99 per month tier provides up to 20 times Pro limits and is where Deep Think and Gemini Spark are available. No official UK Sterling prices for these tiers have been confirmed. The old $249.99 Ultra tier no longer exists following the I/O restructure.

How it fits into the wider picture

Gemini 3.5 Flash — the smaller, faster sibling — has been publicly available since 19 May and already outperforms Gemini 3.1 Pro on coding and agentic benchmarks while running roughly four times faster and at lower cost. The Pro model is intended to push further on the hardest tasks: abstract reasoning, complex multi-step coding, and frontier-level document analysis.

Google's structural advantage remains distribution. Gemini is embedded in Search, Workspace, Android, and — through a reported partnership — Apple's Siri integration. That reach means even a delayed launch carries significant industry weight, irrespective of where the benchmarks eventually land against OpenAI's GPT-5.6 and Anthropic's Fable 5.

What to do right now

• Free users: The improvement you'll feel soonest is through Gemini 3.5 Flash, which is already live.
• Developers and API users: Hold off on migration planning until Google publishes the official model card, confirmed pricing, and availability documentation.
• Ultra subscribers: If you're on the $99.99/month tier, Deep Think access is not guaranteed at launch — the higher $199.99 tier is where that feature is confirmed to sit.
• UK buyers generally: Sterling pricing for subscriptions is unconfirmed. Watch for an official announcement from Google, which has not yet set a public date.

If 17 July holds, the window is five days away. Given the model's track record of slipping, a degree of scepticism is warranted until Google posts something official.