Google has released Gemini 3, which features major improvements in reasoning, multimodal understanding, and more intelligent interactions across Google products. According to the company, it can produce richer visuals and more interactive outputs due to the higher level of understanding.
Tulsee Doshi, Senior Director of Product Management at Gemini, shared that Gemini 3 can now help with learning, creating projects, and planning tasks. He called it an AI that can be a partner, not just a simple answer-machine.
Doshi also said that the model can convert text, images, and other formats, because it understands small details in different types of content. Google engineers have also used it to discover more than 2 million new materials that could help in better batteries, chips, and clean energy tech.
A big highlight is the new Antigravity feature—an agentic coding system where AI can plan, code, and test tasks with virtually no human intervention. It can control the code editor, terminal, and browser for a series of tasks, turning developers into architects.
According to Koray Kavukcuoglu of Google DeepMind, Antigravity has a dedicated agent interface so it can operate AI at a higher level. Gemini 3 is available in Gemini app, AI mode, Search, Gemini API, AI Studio, and for business, in Vertex AI and Gemini Enterprise.
