Announcing Gemini 3: The Most Intelligent New AI Model

AI

Key Points of This Article

• Google announces “Gemini 3,” its new generation of Google AI models.
• It is being rolled out on “Google AI Studio,” the next-generation multimodal generative AI development platform; “Vertex AI,” the machine learning development platform; “Google Antigravity,” the new agent development platform for advanced task-oriented work; and the “Gemini” app.
• Additional AI models are scheduled to be announced soon.

On Tuesday, November 18, 2025, Google announced “Gemini 3,” its new generation of Google AI models.

On the same day as its announcement, “Gemini 3” became available on the next-generation multimodal generative AI development platform “Google AI Studio,” the machine learning development platform “Vertex AI,” the new agent development platform for advanced task-oriented work “Google Antigravity,” and the “Gemini” app. This marks the first time a new generation of Google AI models has been rolled out to general users in the “Gemini” app on the day of its announcement.

“Gemini 3”

“Gemini 3” is a new Google AI model that takes state-of-the-art reasoning and multimodal capabilities to the next level.

In the “Gemini” app, “Gemini 3 Pro (Preview),” an upgraded model of “Gemini 2.5 Pro,” is being rolled out as a new “reasoning model.” Furthermore, for members of “Google AI Ultra,” the top-tier “Google AI” subscription plan on “Google One,” an enhanced reasoning mode based on “Gemini 3” called “Gemini 3 Deep Think,” which features dramatically improved reasoning and multimodal understanding, will also be launched as soon as its safety has been confirmed by testers.

Additionally, for “Google AI Pro” and “Google AI Ultra” members, the base AI model for “AI Mode” has been upgraded to “Gemini 3.” This enables new generative UI features such as immersive visual layouts, interactive tools, and simulations.

In addition, a new agent feature called “Gemini Agent” is being rolled out for “Google AI Ultra” members. It navigates and performs user workflows and other tasks with controlled guidance. It will handle delegated tasks in coordination with Google services, such as organizing an overflowing “Gmail” inbox.

With “Gemini 3,” AI models such as “Gemini 3 Pro,” “Gemini 3 Deep Think,” and “Gemini Agent” have been announced and will be rolled out sequentially. Additional AI models are also scheduled to be announced soon.

Source: Google