Google has released Gemini 3.

Google announced Gemini 3, which it claims marks a new phase in AI. The business touts the model as its smartest yet, capable of realizing user ideas with improved understanding of intent, requiring fewer prompts.

Google subtly mentioned OpenAI in its announcement, noting Gemini 3 Pro gives responses with less unnecessary praise, a known problem with ChatGPT. Google stated in November 2025 that its responses are insightful, concise, and direct, delivering necessary information instead of just what users want to hear.

Gemini 3's capabilities unlock enhancements to Google apps, such as building complex programs within the Gemini app's Canvas workspace. It can recognize handwritten recipes in multiple languages to create shareable cookbooks.

Gemini 3 also enables generative user interface experiences in AI Mode, creating visual layouts with interactive tools and simulations based on user queries.

Google will soon introduce Gemini 3 Deep Think, a research-focused version that will outperform Gemini 3 Pro, already leading in some benchmarks. Gemini 3 Pro scored 37.5 on the Humanity's Last Exam benchmark, surpassing GPT-5 Pro's 31.64. It also leads the LMArena leaderboard, succeeding Gemini 2.5, which Google's prior model held for months, just hours after xAI's Grok 4.1.

Google's CEO, Sundar Pichai, noted the Gemini app has over 650 million monthly active users, with 2 billion using AI Overviews monthly. Over 70% of Google's cloud clients use AI, and over 13 million developers use its models.

The company also launched Google Antigravity, a Gemini-powered coding interface that combines a prompt window, command-line interface, and browser. The app is available in free preview for macOS, Windows, and Linux.

AI Overviews dominate the Search page, and AI Mode is prominently positioned, potentially bothering users less enthusiastic about AI.