Google releases Gemini 3 in push to lead AI race

CHONGQING, CHINA - NOVEMBER 17: In this photo illustration, the logo of Google is displayed on a smartphone screen with the logo of Gemini 3.0 in the background on November 17, 2025 in Beijing, China. Google has announced the development of its next-generation AI model, Gemini 3.0, with a release scheduled for late 2025. (Photo illustration by Li Hongbo/VCG via Getty Images) Gemini 3 includes a powerful new digital "agent" for creating applications on command.

Google has debuted its latest Gemini AI, using it to enhance its dedicated app and popular search engine, as the tech titan competes in a crowded race to dominate artificial intelligence technology.

Gemini 3 is touted by its creators as the “best model in the world” for interpreting and processing data from text, images, audio and video, and as enabling a powerful new digital “agent” for creating applications on command.

“This is our most intelligent model,” Google chief of AI Koray Kavukcuoglu said in a briefing with journalists. “We like to think that it will help everyone bring any idea that they have to life.”

Gemini 3 Pro is “natively multimodal,” Google says, meaning it can process text, images, and audio all at once, rather than handling them separately. As an example, the firm says Gemini 3 Pro could be used to translate photos of recipes and then transform them into a cookbook, or it could create interactive flashcards based on a series of video lectures.

The new AI model will be available in a Gemini app with more than 650 million monthly users and in Google’s world-leading search engine, where Gemini has more than two billion users monthly through its Overviews function, according to Kavukcuoglu.

“Gemini has set quite a new pace in terms of both releasing the models but also getting them to people faster than ever before,” he said. “We have been shipping very rapidly and learning through all that experience.”

Caught off guard by ChatGPT and mocked for early blunders with its own generative artificial intelligence efforts, Google has pulled off a dramatic turnaround since the first generation of Gemini was launched in late 2023, becoming a major player in consumer-facing AI.

Just months later, the California-based giant unveiled AI Overviews, a feature integrated into Google Search that caused online ridicule for its sometimes outlandish answers to queries.

The release of Gemini’s latest iteration, only nine months after the previous version, shows that Google is not letting up in the AI race that has ChatGPT-maker OpenAI fighting hard to maintain its dominance in the generative AI field.

That has seen the big tech companies pump tens of billions of dollars in capital expenditure to pay for AI’s huge costs, with Wall Street’s deep faith in the promise of the AI revolution beginning to show cracks.

Google CEO Sundar Pichai acknowledged to the BBC that there was “irrationality” behind the boom in artificial intelligence investment, which has fuelled a historic rally on the financial markets.

He warned of the “immense” energy requirements of AI and said that people should not “blindly trust” everything AI results tell them, defending the benefit of using other tools like search engines in conjunction with chatbots or overviews.

Google vice-president of product Robby Stein called the integration of Gemini 3 a “huge step forward”.

“What’s possible with Gemini 3 is incredible reasoning capability that you’ve heard so much about,” Stein said in the briefing. “You’ll be able to answer some of your hardest questions right within search.”

Gemini 3 responds to natural language queries or commands with a level of depth and nuance not previously seen, Tulsee Doshi, Google’s senior director of product management, contended.

“The responses that users get from this model are clever, concise, direct, and we try to make sure that it avoids cliche and flattery,” Kavukcuoglu said.

Google also introduced a new software development platform named Antigravity that provides an AI agent capable of autonomously coding as directed in plain language.

The Gemini team said it also ramped up security in the latest model, making it more resistant to cyber attacks, prompt injections and other issues.

– Agence France-Presse