Google's Gemini 3 aims for AI supremacy with PhD-level reasoning

Google has ushered in a new era for its AI ambitions with the launch of Gemini 3 which it calls its “most intelligent model” to date.

Google CEO Sundar Pichai officially made this announcement reflecting a major step in the company's pursuit of advanced AI capable of rivalling human reasoning.

The flagship model, Gemini 3 Pro, is available now in the Gemini app and for the first time the company has integrated it directly into Google Search on day one.

Developers can also access it through AI Studio and Vertex AI. The model is capable of state-of-art reasoning, designed to grasp death and nuance in complex problems.

Google claims it significantly outperforms its predecessor i.e., Gemini 2.5 Pro, across major benchmarks, achieving top scores in areas such as PhD-level reasoning and mathematics.

A major improvement made in this model is a shift away from “flattery” or “cliché.”

Rather, Gemini 3 is designed to provide concise, direct, and genuine insight, acting as a "true thought partner.”

It also allows users to learn complex topics via interactive visualizations, build richer application by following “vibe coding,” and plan multi-step tasks with enhanced long-horizon planning.

In addition to the standard Pro model, Google introduced “Gemini 3 Deep Think,” an improved reasoning model that gives even more improved results.

Specifically for developers, the company has unveiled a new agentic development platform named “Google Antigravity” which uses Gemini 3 to autonomously plan and execute complex software tasks.

The company also claims it to be its most secure model yet citing that Gemini 3 has undergone extensive safety evaluations.

The rollout starts immediately, marking the next chapter in Google’s rapidly accelerating AI roadmap.