Google launches Gemini 3, integrating multimodal AI and intelligent agent features

According to ChainCatcher, the official Gemini account @GeminiApp announced that Google has officially launched its new generation AI model, Gemini 3, which features a million-token context window, multimodal understanding capabilities, and supports unified analysis of video, audio, images, and text.

The new version introduces Gemini Agent, which can complete multi-step tasks under user guidance and is currently open for testing to Ultra users in the United States. Gemini 3 has been launched globally, and student users can try the Pro plan for free for one year.