‘Our most advanced AI model’ — Google launches Gemini 3

Google has launched Gemini 3, an artificial intelligence (AI) tool, described as the company’s most advanced AI model.

The company announced the launch on Tuesday, describing the model as a significant leap in intelligence and reasoning, introducing a new generation of agentic AI across its products.

Sundar Pichai, chief executive officer (CEO) of Google and Alphabet, said in this “new chapter, we’ll continue to push the frontiers of intelligence, agents, and personalisation”.

Google noted that global interest in AI is rapidly rising, with Nigeria recording over a 200 percent increase in Google search queries for “AI and studying,” and a 160 percent surge in searches related to “how to learn AI” in the past year.

The company said the Gemini 3 family debuts with Gemini 3 Pro, which Google says demonstrates deeper comprehension and advanced reasoning abilities.

According to Google, the model sets new performance records across multiple industry benchmarks, scoring 91.9 percent on the GPQA Diamond test for complex problem-solving and excelling at interpreting images, videos, and handwritten text.

The company said Gemini 3 Pro delivers more precise and insightful responses, designed to serve as a collaborative partner rather than a generic assistant.

Google also highlighted the model’s improved ability to perform multi-step tasks, understand user intent, and get the job done with less user prompting. 

Also, the tech giant said it likewise enables developers to convert ideas into functional code with ease, thereby excelling at “vibe coding”.

According to the company, the technology is now integrated into Google search, offering a more interactive and learning-focused experience.

Google said Gemini 3 automatically generates dynamic visual layouts, simulations, and tools tailored to a user’s query.

The company added that users exploring physics concepts can engage with a three-body problem simulation, while those researching mortgage options can access a custom loan calculator built in real time.

Google also introduced an upgraded agent in the Gemini app that can execute tasks across Google’s ecosystem.

For Google AI Ultra subscribers, the company said the experimental Gemini Agent can organise emails, draft responses, and complete multi-step tasks such as finding travel details and comparing rental options.

In addition, the tech giant said a user can request it to “organise my inbox,” and the agent will prioritise tasks and prepare replies, while another prompt like “help me book a mid-size SUV for my trip” will allow it to pull flight information, compare rental cars, and prepare the booking.

GOOGLE LAUNCHES TOOL TO SUPPORT DEVELOPERS

To support developers, Google launched Google Antigravity, a new agentic development platform designed to plan and complete software tasks autonomously.

The company described the platform as a “mission control” system for managing multiple AI agents, ensuring transparency, verifiable outputs, and developer oversight.

Google said this marks a shift toward an agent-first approach to software development, enabling creators to focus on strategy while AI handles complex execution.

The company added that Gemini 3 underwent its most extensive safety reviews to date.

Google confirmed that Gemini 3 Pro is now available in the Gemini app and can also be accessed by Google AI Pro and Ultra subscribers through AI Mode in search.