Overview
- Gemini 3 Pro is live in the Gemini app for everyone today, with Search's AI Mode access starting for AI Pro and Ultra subscribers in the US.
- The model leads LMArena with a 1501 score and posts 37.5% on Humanity's Last Exam, with Google also citing gains on multimodal tests like MMMU‑Pro and Video‑MMMU.
- Generative interfaces debut as Dynamic View and visual layouts that can code custom tools, simulations, and magazine‑style responses tailored to each prompt.
- An experimental Gemini Agent can carry out multi‑step tasks in Google apps such as organizing Gmail or researching and booking travel, with availability limited to paying tiers during rollout.
- Developers and enterprises can use Gemini 3 via the Gemini API, AI Studio, Vertex AI, the Gemini CLI, and the new Antigravity agentic IDE, with availability also confirmed in Canada.