Google Launches Gemini 3, Its Most Advanced AI Model, Setting New Benchmarks in Reasoning

Google has introduced Gemini 3, its most powerful foundation model to date, now available through the Gemini app and AI Search. Arriving just seven months after Gemini 2.5, the new model positions Google at the forefront of frontier AI development amid rapid releases from OpenAI and Anthropic. A more research-focused variant, Gemini 3 Deepthink, will roll out to Google AI Ultra subscribers after completing additional safety evaluations.

Led by Tulsee Doshi, Google’s product head for Gemini, the company highlighted significant gains in reasoning and depth of response — improvements reflected in independent benchmarks. Gemini 3 set a record score on the Humanity’s Last Exam and climbed to the top of the LMArena leaderboard. With over 650 million monthly users and 13 million developers leveraging the ecosystem, Google also released Google Antigravity, a new Gemini-powered multi-pane coding interface that blends chat, terminal, and browser views to support advanced agentic coding.