Google Launches Gemini 3, Its Most Advanced AI Model with Major Breakthroughs in Reasoning & Coding

Google has unveiled Gemini 3, its most powerful foundation model to date, now available through the Gemini app and AI-powered search. Arriving just seven months after Gemini 2.5—and shortly after rival releases like GPT 5.1 and Anthropic’s Sonnet 4.5—the model highlights Google’s accelerated push to lead the AI frontier. A more advanced variant, Gemini 3 Deepthink, aimed at research and high-level reasoning tasks, will soon be released to Google AI Ultra subscribers following additional safety checks.

Google says Gemini 3 delivers unprecedented improvements in reasoning, depth and nuanced responses, with internal benchmarks supporting the claims. The model achieved a top score of 37.4 on the Humanity’s Last Exam benchmark, surpassing GPT 5 Pro, and also claimed the No.1 spot on the user-driven LMArena leaderboard. With more than 650 million monthly users of the Gemini app and 13 million developers relying on the system, Google continues to strengthen its AI ecosystem.

Alongside the model, Google introduced Antigravity, a new coding environment powered by Gemini. The tool provides multi-pane agentic coding through an integrated prompt window, terminal and browser view, allowing the model to write, test and refine code in real time. With Gemini 3 leading multiple benchmarks and Deepthink on the horizon, Google is signalling its determination to stay ahead in the rapidly intensifying global AI race.

Pic Courtesy: google/ images are subject to copyright