Google launches Gemini 3 AI model with agent features

Code Mastery Centre
18 Nov, 2025

Gemini 3 AI

Google unveiled Gemini 3 on November 18, introducing what the company describes as its most advanced artificial intelligence model to date, featuring enhanced reasoning capabilities and autonomous agent functions that can manage complex, multi-step tasks.​

The new model is immediately available to all users through the Gemini app, marking the first time Google has provided universal access to its premier AI system on launch day. Gemini 3 Pro currently ranks at the top of the LMSYS leaderboard, a widely used platform for evaluating AI models.

Agent Capabilities and Generative Interfaces

Gemini 3 introduces what Google calls "generative interfaces," allowing the model to autonomously determine the most suitable output format for queries rather than defaulting to text. When users request travel suggestions, for instance, the model can generate website-like interfaces featuring images, modules, and interactive follow-up questions.​

The experimental Gemini Agent feature, available to Google AI Pro and Ultra subscribers in the United States, can connect to services like Calendar, Gmail, and Reminders to perform tasks such as organizing inboxes or coordinating schedules. According to Tulsee Doshi, senior director and head of product at Google DeepMind, the model provides "intelligent, succinct, and direct answers, replacing clichés and flattery with authentic insights".

New Development Platform and Shopping Integration

Alongside Gemini 3, Google launched Antigravity, an agent-first development platform that enables multiple agents to access editors, terminals, and browsers simultaneously. The platform generates self-documenting "Artifacts" including task lists, strategies, and screenshots to validate work performed.​

Google also integrated Gemini 3 with its Shopping Graph, which contains over 50 billion product listings updated hourly, allowing users to receive interactive product recommendations directly within the app. The enhanced reasoning mode, Gemini 3 Deep Think, is currently available to safety testers before broader rollout to Ultra subscribers. 

Next Post Previous Post