Overview
- Gemini 3 Pro is live today in the Gemini app for all users, with paid Google AI Pro and Ultra subscribers gaining access in Search's AI Mode via the new Thinking option and broader Search availability to follow.
- Google touts major gains in reasoning, multimodal understanding, and factuality, with Gemini 3 Pro topping LMArena at 1501 and showing reduced sycophancy and stronger safety protections.
- A new Deep Think mode further boosts reasoning and is limited to safety testers initially, with availability to AI Ultra subscribers planned in the coming weeks.
- Generative interfaces debut as experiments, producing dynamic, magazine-style layouts and on-the-fly tools and simulations in the Gemini app and in Search responses.
- Agentic capabilities expand with the experimental Gemini Agent for multi-step tasks and the Antigravity development platform, now in public preview on Mac, Windows, and Linux, alongside developer access through AI Studio, Vertex AI, and the Gemini API/CLI.