Gemini 3: Google Introduces Company's 'Most Advanced Multimodal AI Model' To Date

Google has launched Gemini 3, its most advanced AI model to date, with CEO Sundar Pichai declaring it "the best model in the world for multimodal understanding." The announcement was made Tuesday, via a post on X, where Pichai highlighted the model's enhanced agentic and coding capabilities, stating it can "bring any idea to life" by quickly grasping context and intent with less prompting.

Gemini 3 is described as the company's most intelligent model yet, combining state-of-the-art reasoning, multimodal understanding, and a 1 million-token context window. It excels at interpreting complex information, such as translating handwritten recipes across languages, generating interactive learning materials, and analyzing sports performance videos.

The model achieves top scores on major benchmarks, including a groundbreaking 1501 Elo on the LMArena Leaderboard, 37.5% on Humanity’s Last Exam, 91.9% on GPQA Diamond, and 87.6% on Video-MMMU. It also leads the WebDev Arena leaderboard and shows strong performance in coding-agent tests.

Google DeepMind CEO Demis Hassabis emphasized that Gemini 3 represents a significant step toward artificial general intelligence (AGI), with the model capable of solving complex problems in science and mathematics reliably. The model is also designed to act as a "true thought partner," offering genuine insight rather than clichéd responses.

A specialized reasoning mode, Gemini 3 Deep Think, is currently in preview for safety testers and will eventually be available to Google AI Ultra subscribers. This mode is intended for tackling highly complex analytical tasks with enhanced tool-use abilities.

For developers, Google introduced Google Antigravity, a new agentic development platform enabling AI agents to autonomously plan and execute workflows across code editors, terminals, and browsers.

The launch comes as Google seeks to strengthen its position in the competitive AI landscape, following the release of rival models like xAI's Grok 4.1, OpenAI’s GPT-5.

Gemini 3 Pro is being released in preview across Google's ecosystem and products, including the Gemini app, AI Mode in Search, and developer platforms like AI Studio and Vertex AI, with full availability starting today.