Alphabet’s Google on Tuesday introduced the newest version of its artificial intelligence model, Gemini 3, saying that new features would be immediately available in a handful of money-making products, including its search engine.
And the arrival of Gemini 3, on paper, looks set to keep Google at the head of the AI race, albeit 11 months after version 2. Executives emphasized in a press briefing that Gemini 3 held the top spot on several popular industry leaderboards for AI model performance.
“It’s our most advanced model,” CEO Sundar Pichai wrote in a company blog post.
But the AI race has been driven less by benchmarks and more by money-making applications of the technology, as Wall Street questions whether an AI bubble is forming. Alphabet’s stock has been lifted this year and last by the financial success of AI offerings from its cloud computing division.
But even with heavyweights like Google Ai, OpenAI, and Anthropic backing them, new AI model updates have struggled to stand out from one another, capturing attention mainly when they fail—an experience Meta faced earlier this year.
Google pointed out that, rather than depending on retrospective launches, Gemini 3 already supported several in-market revenue-generating consumer and enterprise products at launch.
“We believe that Gemini sets, both in terms of releasing the model and getting it out for people to use faster than anything we’ve ever released.” Koray Kavukcuoglu, Google’s chief AI architect, said to reporters in the run-up to today’s briefing.
With the launch of Gemini 3, Google began using its so-called F.T.C.-approved model from the start, Pichai said. In the past, new releases of Gemini could take weeks or months to be incorporated into Google’s most heavily used products.
For paying users of Google’s premium AI subscription service, subscribers will receive Gemini 3 features in AI Mode, a search function that filters out the usual chaff in favor of computer-generated answers for difficult questions.
New Features
Modifications to Gemini 3 in areas such as coding and reasoning allowed Google to develop a suite of new capabilities that serve consumers and enterprises alike.
(On Wednesday, the company introduced “Gemini Agent,” a feature that can handle multistep tasks, such as organizing a user’s inbox or booking travel.) The tool is a step towards what Google’s AI chief, Demis Hassabis, said last year would be the ultimate AI tool — a “universal assistant” known internally at the company as AlphaAssist, Reuters has previously reported.
Google also overhauled the Gemini app — released for Android and iOS devices last year — so it returned results akin to a full website, hitting content publishers who rely on web traffic to make money.
Josh Woodward, the app’s vice president, showed reporters how Gemini can now answer a query like “create a Van Gogh gallery with life context for each piece” by creating, via an on-demand interface, a visual, interactive result.
For business customers, Google showcased a new product it plans to release soon called Antigravity, a software development platform that enables AI agents to plan and complete coding tasks on their own.
