Gemini 3.5 Pro: Feed a 2M-Token Codebase in One Call

Google started rolling Gemini 3.5 Pro into general availability in the last week of June 2026, and the headline spec is the one every developer fixated on: a 2,000,000-token context window, the largest of any production frontier model right now. June 2026 is already being called the biggest model-launch month ever, and this is the release that actually changes how you architect an app.

Two million tokens is not a bigger version of the same thing. It is a different design point. A single request can now hold an entire mid-size codebase, several full SEC filings, or years of chat history without chunking, embeddings, or a vector database. The paradigm the Gemini docs push is blunt: instead of retrieving the relevant slice, just put all the relevant information in the prompt and let the model reason over it.