Gemini 3 launch gives Google an edge in the fiercely competitive AI race

With the launch of Gemini 3, the next major version of the company’s frontier model, Google aims to push ahead to the forefront of the fiercely competitive AI race.

Gemini 3 will also be central for the tech giant’s efforts to maintain the growing momentum in its cloud unit, a crucial part of Google’s strategy to monetise its AI offerings.

In recent quarters, Google Cloud has emerged as one of the firm's fastest-growing segments, benefiting from surging demand from enterprises for AI products and services. It was also one of the key drivers for Google parent Alphabet to post its first-ever $100 billion quarter in Q3̣ 2025.

Google CEO Sundar Pichai said that more than 70 percent of the company’s Cloud customers use their AI offerings and 13 million developers have used its generative models for development.

How is Google releasing new AI models so quickly?

The launch comes less than a year after Google introduced Gemini 2 and about seven months after the Gemini 2.5 release, underscoring the rapid pace at which the frontier models are evolving at the tech giant.

Pichai said the company's ability to get advanced capabilities to the world faster than ever is a direct result of its full-stack approach to AI, wherein it is investing in developing every layer of the AI technology stack: Research, models and tooling, AI infrastructure, and product and platforms that reach billions of users.

"It's amazing to think that in just two years, Al has evolved from simply reading text and images to reading the room," he said.

What's new in Gemini 3?

Google states that Gemini 3 is its most intelligent model, which builds upon and combines all of Gemini's predecessors' capabilities. This includes Gemini 1's native multimodality and understanding vast amounts of information and Gemini 2's agentic capabilities and an expansion of its reasoning and thinking capabilities.

Gemini 3 is able to grasp more depth and nuance, whether that means picking up subtle clues in a creative idea or peeling apart the overlapping layers of a difficult problem and provide clear, direct answers.

"(Gemini 3)'s responses are smart, concise and direct, trading cliché and flattery for genuine insight — telling you what you need to hear, not just what you want to hear," Google DeepMind CEO Demis Hassabis and CTO Koray Kavukcuoglu said in a blogpost.

The executives also note that Gemini 3 is their most factually accurate model, making it capable of more reliably solving complex problems across a range of topics like science and mathematics.

Google stated that the information produced by Gemini 3 was 72 percent accurate, in comparison to Gemini 2.5 Pro's 54.5 percent and GPT 5.1's 34.9 percent, according to a standard benchmark test. It also noted that the model scores better than other frontier models on several other benchmarks.

Gemini 3 is the strongest model in the world for multimodal understanding and reasoning, Hassabis and Kavukcuoglu said in the blogpost.

This means that the model can take anything - from text to images and videos to code - and respond in a format that suits the user's preferences. For instance, text lessons can be converted into visual flashcards, an image can be turned into a board game, analysing long-form sports videos to identify technical issues and provide exercises to correct them.

How will Gemini 3 be available to consumers and developers?

Google is currently releasing Gemini 3 Pro in preview, making it available across a suite of its products.

It will be available to consumers across the world through the Gemini app, with higher limits for Google's AI Plus, Pro and Ultra subscribers. The AI Plus tier is not available in India, but Google is currently offering the AI Pro tier to Jio customers for free for 18 months. Earlier this year, it offered students a free one-year subscription to this plan.

For developers, the model will be available through AI Studio, Gemini CLI, and its newly launched agentic development platform, Google Antigravity. It will also be available through third-party platforms such as Cursor, GitHub, JetBrains, Manus, and Replit among others.

Enterprises will get access to Gemini 3 through Vertex AI and the recently-introduced Gemini Enterprise.

Will Google offer local data processing for Gemini 3 in India?

Saurabh Tiwary, Vice President and General Manager of Cloud AI at Google, said last week that the company is opening early testing of its latest and most advanced Gemini models to customers in India.

He added that Google is committed to launching its most powerful Gemini models in the country with full data residency support. However, the company is yet to disclose a specific timeline for the Gemini 3.

What changes are coming to Gemini app?

Gemini 3 has enabled a new capability called 'generative interfaces', that allows the model to create visually immersive and interactive user experiences for any user prompt on the fly, the company said.

The first two experiments that are rolling out as part of this initiative are visual layout and dynamic view.

Visual layout creates an immersive, magazine-style format with photos and modules that one can interact with to further customise their response while dynamic view creates a custom user interface tailored to the user's prompt.

Gemini app has over 650 million monthly active users, Pichai said in October, up from 350 million in March. In comparison, OpenAI recently stated that ChatGPT has over 800 million weekly users, while Meta AI is now used by more than a billion monthly active users across Meta’s apps.

What can Gemini Agent do?

Google is introducing an experimental feature called Gemini Agent that can perform multi-step tasks within Gemini, such as organising email inbox, managing calendar, adding reminders, researching and booking travel among others. It will take confirmation before critical actions like making purchases or sending messages, the company stated. The feature is currently available to Google AI Ultra subscribers in the United States.

In addition, people can now build more full-featured web apps through the Canvas feature. The responses are more helpful, better formatted and more concise, the company said.

How is Gemini 3 improving Google Search?

Gemini 3 will power Al Mode, Google's AI-powered answer engine-style search experience, offering more complex reasoning and generative user interfaces with dynamic visual layouts.

"This is the first time we are shipping Gemini in Search on day one," Pichai said.

Google Search's query fan-out technique, currently used by AI Mode, is getting a major upgrade due to Gemini 3's advanced reasoning, according to Google's search chief Elizabeth Reid.

"Not only can it perform even more searches to uncover relevant web content, but because Gemini more intelligently understands your intent it can find new content that it may have previously missed. This means Search can help you find even more credible, highly relevant content for your specific question," Reid said.

This feature will initially be available to Google AI Pro and Ultra subscribers in the United States with a broader rollout expected in the forthcoming future.

How is Gemini 3 fueling agentic coding?

Google unveiled a new agentic development platform called Google Antigravity that aims to rethink the traditional Integrated Development Environment (IDE) for the agent-first era

Powered by Gemini 3 Pro, it helps developers build faster by collaborating with intelligent agents that can operate autonomously across the code editor, terminal, and browser. These agents can plan and execute complex, end-to-end software tasks simultaneously on the developer’s behalf while validating their own code.

To enable this, Google Antigravity comes bundled with the Gemini 2.5 Computer Use model for browser control and the Nano Banana image editing model, known as Gemini 2.5 Image. It also supports Anthropic’s Claude Sonnet 4.5 models, and OpenAI’s GPT-OSS.

This launch comes after Google struck a $2.4 billion deal in July to hire Varun Mohan, CEO of agentic coding startup Windsurf along with other research and development employees and to take a nonexclusive license to certain Windsurf technology. Windsurf was later acquired by Cognition.

Antigravity is currently available in public preview on MacOS, Windows and Linux.