Google launches Gemini 3 across Search and developer tools

Google launched Gemini 3, describing it as its “most intelligent model” and positioning it as a stronger contender among top AI systems. The new LLM is designed to improve reasoning, understand more nuance in tasks, and better infer the context and intent behind user requests, which Google says should help people get things done with fewer prompts.

Gemini 3 is already being rolled out across Google products. It is live in AI mode in Google Search, marking the first time Google has shipped a new model in Search on launch day, though access is currently limited to Google AI Pro and Ultra subscribers. The model is also available in the Gemini app, AI Studio, Vertex AI, and Google’s agentic development platform, Antigravity.

Google highlighted benchmark results for Gemini 3 Pro, which leads LMArena with a score of 1501 Elo, ahead of xAI’s Grok and Gemini 2.5 Pro. Gemini 3 Deep Think is aimed at more complex tasks with stronger reasoning, including performance gains on Humanity’s Last Exam. The Gemini app also gained a My Stuff folder, shopping improvements, generative interfaces, and a Gemini Agent feature for tasks such as appointments, reminders, inbox organization, and online research.