Google has officially launched Gemini 3, its most advanced artificial intelligence model yet, calling it a major leap in reasoning, understanding, and overall performance. The new model replaces Gemini 2.5 and is now available across Google Search’s AI Mode, the Gemini app, Chrome, and Google Cloud.
According to the company, Gemini 3 delivers smarter responses with better context awareness, helping users get accurate answers without needing long or complicated prompts. Google CEO Sundar Pichai said the model shows big progress in understanding user intent, making interactions more natural and useful.
Google also claims Gemini 3 is now the strongest multimodal model in the world, able to handle text, images, code, and visual content at a deeper and more advanced level than past versions.
Gemini 3 Pro arrives first
The rollout begins with Gemini 3 Pro, available in preview. Google says it performs better than Gemini 2.5 Pro on every major benchmark, offering clearer, more thoughtful answers packed with detail.
Another version, Gemini 3 Deep Think, is still being tested for safety. It will later be released for Google AI Ultra subscribers and is designed for tasks that require long, detailed reasoning.
Stronger reasoning, coding, and analysis
Gemini 3 introduces major improvements in logic and understanding. The model can now convert information across formats, create apps and visualizations from a single prompt, solve advanced math and science problems, and provide interactive visual answers directly in Google Search. It also improves coding support, offering deeper workflows for developers.
Google DeepMind CTO Koray Kavukcuoglu described Gemini 3 as the company’s “most intelligent model,” capable of helping people bring complex ideas to life.
Developer platform Antigravity
Along with Gemini 3, Google introduced Antigravity, a new platform that supports autonomous AI coding agents. These agents can write, test, and verify code across different tools, almost like automated software engineers. In a demo, Google showed an AI agent building a full flight-tracking website on its own.
Gemini 3 Deep Think takes a different approach by testing multiple reasoning paths before choosing the most accurate answer. This version is tailored for detailed work such as research, strategic planning, and long coding problems. It will be available only to Google AI Ultra subscribers at $249.99 per month.
Google says safety has been heavily strengthened. Gemini 3 was trained to resist prompt injection attacks and prevent harmful or misleading outputs. This improvement comes as Google increases AI features across its main products.
Gemini’s expansion is also helping Google financially. Last quarter, Google Cloud revenue grew 34% year-over-year to $15.2 billion, with nearly 70% of cloud customers using Google’s AI tools. Company executives say Gemini has become one of the biggest drivers of growth.
As part of its wider rollout, Google will also give U.S. college students free access to Gemini Pro for one year.
