Google introduced Gemini 3 as its most intelligent AI model yet, aimed at helping users learn, build and plan virtually anything. It brings major improvements in reasoning, multimodal understanding (text, images, video, code) and long-context capabilities, and is being made available across Google’s suite of products including Search (AI Mode), the Gemini app, AI Studio, Vertex AI and the new agent platform “Google Antigravity”. The model leads in numerous benchmark tests, with a preview of an even stronger “Deep Think” mode slated for release to Ultra subscribers soon. The company also emphasized that Gemini 3 is the most safety-evaluated model in its lineup, with extensive internal and external reviews to reduce misuse, biases and vulnerabilities.
