Google has released Gemini 3, its most advanced foundation model to date, positioning the system as a major leap in reasoning, coding assistance and real-world task handling. The launch comes just seven months after Gemini 2.5 and arrives amid rapid competition across the AI industry, following recent releases from OpenAI and Anthropic.
The model is now available through the Gemini app and Google’s AI-powered search interface.
Google also announced Gemini 3 Deepthink, a research-optimized version designed for deeper reasoning and complex analysis. Deepthink will be available to Google AI Ultra subscribers once additional safety evaluations conclude.
“With Gemini 3, we’re seeing a massive jump in reasoning,” said Tulsee Doshi, Google’s head of product for Gemini. “It’s responding with a level of depth and nuance that we haven’t seen before.”
Early results back those claims. Gemini 3 set a new record on the Humanity’s Last Exam benchmark with a score of 37.4, surpassing the previous high of 31.64 held by GPT-5 Pro. It also ranked first on LMArena, a human-evaluated benchmark that measures user satisfaction and conversational quality.
Google reports the Gemini app now has more than 650 million monthly active users, while more than 13 million developers have incorporated Gemini tools into their workflows.
Alongside the new model, Google introduced Google Antigravity, a Gemini-powered coding interface that blends multi-pane editing with AI-assisted development. The tool integrates a prompt window, command-line environment and an in-browser preview system designed to update in real time as code changes.
“The agent can work with your editor, your terminal and your browser to help you build applications in the best way possible,” said DeepMind CTO Koray Kavukcuoglu.
Google says Gemini 3’s release marks the start of a new rollout phase focused on deeper reasoning, multimodal capabilities and expanded developer tools.