Google began rolling out Gemini 3.5 Pro on Friday, a long-anticipated update that pairs a two-million-token context window with a more deliberate "Deep Think" reasoning mode, escalating a widening price war over AI inference.

The release lands months after rival launches from OpenAI and xAI, whose GPT-5.6 series and Grok 4.5 set the stage for a summer of competitive model debuts. Google is betting that raw context length and stronger reasoning will justify its positioning against well-funded challengers.

A two-million-token window allows the model to ingest hundreds of pages of documents, entire codebases or lengthy meeting transcripts in a single pass, a capability Google says is aimed at enterprises processing large volumes of unstructured data. The "Deep Think" mode, reserved for higher service tiers, trades speed for more careful step-by-step problem solving.

Pricing is central to the strategy. Google has set API rates it describes as competitive, hoping to win developers at a moment when the cost of running models — not just training them — has become the decisive battleground for adoption. Rivals have moved in the same direction, squeezing margins across the sector.

The launch also sharpens scrutiny of AI spending. Some analysts have compared the buildup to the dot-com era, warning that returns on enormous data-center investments remain unproven even as model quality improves. Google's challenge is to convert capability into durable revenue.

For now, Gemini 3.5 Pro signals how the frontier has shifted from novelty to infrastructure. Whether customers pay for two-million-token windows at scale will determine whether the latest salvo in the model wars reshapes the market or merely adds to the noise.

Image source: i.ibb.co