Alphabet’s Google on Tuesday unveiled Gemini 3, its most advanced artificial intelligence model to date, and moved to deploy it immediately across key revenue-generating products, including its flagship search engine, in a bid to strengthen its position in the intensifying AI race.
The release comes 11 months after the launch of Gemini 2 and represents Google’s fastest turnaround yet for a major model upgrade. Executives said during a press briefing that Gemini 3 now tops several closely watched industry benchmarks that gauge an AI system’s reasoning, comprehension and problem-solving abilities.
Chief Executive Sundar Pichai called Gemini 3 “our most intelligent model,” adding in a blog post that the rollout reflects the company’s push to accelerate the delivery of new AI capabilities to consumers and businesses. But Wall Street’s focus has shifted in recent months from benchmark performance to whether tech giants can convert advances in AI into sustainable revenue streams, amid growing concerns of a broad AI valuation bubble.
So far, Alphabet’s shares have been supported by rising demand for AI-enhanced services within its cloud division. Even so, analysts say the arms race between Google, OpenAI and Anthropic has made it increasingly difficult for new models to stand out unless they stumble. Meta faced such scrutiny earlier this year after glitches in some of its model releases.
Google sought to counter that narrative on Tuesday, stressing that Gemini 3 is already powering several commercial products at launch—a departure from earlier cycles, when integrating a new model into widely used services could take weeks or months. “We think Gemini has set quite a new pace in terms of releasing models and getting them to people faster than ever before,” said Koray Kavukcuoglu, Google’s chief AI architect.
The integration into Google Search marks the first time a new Gemini model has been embedded from day one. Paying subscribers to Google’s premium AI plan will also gain access to the model through AI Mode, a search option that prioritizes computer-generated answers for complex or multi-layered queries.
Gemini 3’s enhancements in coding, reasoning and planning have allowed Google to introduce a suite of new tools aimed at both consumers and enterprise clients. Among them is “Gemini Agent,” a feature capable of handling multi-step tasks such as organizing email inboxes, assembling travel itineraries or managing administrative workflows. The tool moves Google closer to the “universal assistant” concept that AI chief Demis Hassabis has been championing internally under the codename AlphaAssist, as previously reported by Reuters.
Google also overhauled the Gemini app, giving it the ability to produce responses that resemble full web pages—complete with structured layouts, visuals and interactive elements. Josh Woodward, the vice president overseeing the app, demonstrated how Gemini can generate an instant virtual gallery when asked to “create a Van Gogh gallery with life context for each piece,” displaying artwork, descriptions and navigation tools on demand.
The redesign underscores the mounting tension between AI platforms and online publishers who rely on search-driven traffic, as more user queries are satisfied directly within AI-generated environments.
更多精彩内容，关注钛媒体微信号 （ID：taimeiti），或者下载钛媒体 App
