Google has introduced Gemini 3, its newest and most powerful foundation model, now available through the Gemini app and the company’s AI search interface. Arriving just seven months after Gemini 2.5, the update positions Google at the forefront of the AI race, landing barely a week after OpenAI launched GPT-5.1 and shortly after Anthropic released Sonnet 4.5. Google also announced that a more intensive version, Gemini 3 Deepthink, will roll out soon to AI Ultra subscribers pending additional safety reviews.
Early signs suggest a significant leap in reasoning capabilities. According to Google’s head of product for Gemini, Tulsee Doshi, the model delivers deeper, more nuanced responses than any previous iteration. Independent benchmarks appear to support that claim: Gemini 3 scored 37.4 on the Humanity’s Last Exam benchmark, the highest recorded surpassing GPT-5 Pro’s previous best of 31.64. It also topped the LMArena leaderboard, which measures real user satisfaction.
The rapid growth of the Gemini ecosystem provides a strong foundation for the model’s adoption. Google reports over 650 million monthly active users on the Gemini app and more than 13 million developers integrating the model into their workflows. These numbers highlight the increasing reliance on AI-powered tools across both consumer and enterprise environments.
In addition to the model upgrade, Google unveiled Antigravity, a new coding interface powered by Gemini. The tool blends a chat-style prompt window with a command-line interface and an embedded browser, enabling developers to build and test software in a multi-pane, AI-driven environment. DeepMind CTO Koray Kavukcuoglu says the agentic system is designed to seamlessly assist across the editor, terminal, and browser, offering a more fluid and collaborative coding experience.