Google Launches Gemini 3, its Latest AI Model

Settle down, folks — there’s a new AI model in town. Google announced Gemini 3, less than a year after the debut of its predecessor, Andrew Nusca reported for Fortune Tech.

G3 is better at coding and reasoning than its predecessor. (Image: Google)

CEO Sundar Pichai called G3 — if I may — the company’s “most intelligent model,” though as the industry moves away from chest-thumping about benchmarks, there’s no universal definition of what that means.

G3 is better at coding and reasoning than its predecessor.

Better still, it will be embedded in Google’s lucrative search engine from launch — a first but major step for a company that has promised to spend $85 billion this year on AI-related capital expenditures and even more in 2026.

The launch of Gemini 3 comes with accoutrements, if you will.

There’s a Gemini Agent that promises multi-step task completion. There’s also a redesigned Gemini app that returns more complete answers. And we mustn’t forget the new agent-centric software development platform for businesses, dubbed Antigravity.