Gemini 3.5 Pro, the top-tier AI model Google unveiled at Google I/O in May, is now in the final stretch toward general availability targeted for June. Built around an enormous 2-million-token context window and a new reasoning mode called Deep Think that favors deliberate thinking over speed, it is Google's bid to compete at the frontier of reasoning and multimodal AI. For now, though, it remains in limited preview and has not shipped broadly.
The flagship of the Gemini 3.5 family
Gemini 3.5 Pro sits at the top of Google's latest Gemini 3.5 lineup. It takes over the workloads previously routed to the company's "Ultra" tier — the hardest reasoning, demanding multimodal tasks, and very long context.
The headline specification is the 2-million-token context window. Context length governs how much material a model can weigh at once, and this figure is double the 1 million tokens offered by the lower-tier Flash. A window that large lets Gemini 3.5 Pro hold entire long documents, large codebases, and extended conversations in working memory, widening the range of material it can handle.
The other highlight is the reasoning mode called Deep Think. Rather than answering quickly, it is designed to spend more time working through complex problems, and combined with multimodal understanding across images and text, it aims to perform on the most demanding tasks.
A two-model strategy, and the question of price
Google already shipped the faster, cheaper Gemini 3.5 Flash earlier in the spring. Flash showed stronger coding and agentic performance than the previous generation's Pro, but it gave ground on the hardest reasoning — precisely the gap the new Pro is meant to fill. Pairing Flash for high-volume work with Pro for the heavy lifting is a design common across the industry.
On pricing, the model is expected to arrive first through consumer subscriptions. The entry points are the 20 USD (about 3,200 yen) per month Pro plan and the 250 USD (about 40,000 yen) per month Ultra plan, with Deep Think offered as an Ultra-only feature. Pay-as-you-go API pricing is expected to follow the same ratio as past generations — roughly ten times Flash — at around 15 USD (about 2,400 yen) per million input tokens and 60 USD (about 9,700 yen) per million output tokens. At that level it would compete squarely with frontier models from Anthropic and OpenAI.
※1 USD = 161 JPY (as of June 24, 2026). API pricing figures are estimates.
Still a promise, amid intensifying competition
It is worth noting that, even in late June, Gemini 3.5 Pro has not shipped widely. At I/O, Google CEO Sundar Pichai told the audience, in effect, to wait another month — a line that reportedly drew groans from a crowd hoping for immediate access. The model is running in internal use and in a limited enterprise preview, but broad availability is still pending.
Until it is generally available and independent testers can evaluate it, the capabilities described here remain Google's claims. Launch timelines can slip, so "June" is best read as a target rather than a guarantee. Frontier AI competition is heating up, with rivals shipping capable models in quick succession and aggressively priced Chinese labs adding to the downward pressure on prices. Google's approach has leaned on breadth and distribution — offering models across price points and weaving them into its own products and cloud — rather than betting everything on a single best model. Gemini 3.5 Pro is the piece aimed at the very top of that market.
Summary
With its 2-million-token context and Deep Think mode, Gemini 3.5 Pro is positioned as the flagship of the Gemini family, targeting general availability within June. The first things to watch are whether it ships on schedule and how it holds up once independent reviewers can test it. If it delivers on its specifications, Google gains a credible flagship to set against its rivals' best — and a chance to shed the perception that it announces ahead of shipping.
