Google has officially launched Gemini 3 and embedded it directly into Google Search, a first for its AI models.
According to Elizabeth Hamon Reid, Google’s VP of Engineering for Search, “Gemini 3 brings incredible reasoning power … built to grasp unprecedented depth and nuance for your hardest questions.” She added that this model “unlocks new generative UI experiences … with interactive tools and simulations … generated specifically for you.”
At a press briefing, Google’s chief AI architect, Koray Kavukcuoglu, said they “think Gemini has set quite a new pace … in both releasing the models, but also getting it to people faster than ever before.” Meanwhile, CEO Sundar Pichai described Gemini 3 as “our most intelligent model,” emphasizing the leap in its reasoning and multimodal understanding.
On performance, Google claims the model outperforms Gemini 2.5 Pro on major reasoning tasks and academic benchmarks, even reaching what they characterize as “PhD-level” reasoning in logic, math, and science. They also highlight that Gemini 3 has passed “the most comprehensive set of safety evaluations of any Google AI model to date,” showing “reduced sycophancy, increased resistance to prompt injections and improved protection against misuse via cyberattacks.”
The launch isn’t just symbolic: Gemini 3 is available in Google Search from day one, initially for U.S. users on Google AI Pro and Ultra plans via the “AI Mode.” Google says that thanks to Gemini 3’s upgraded “query fan‑out” technique, Search can issue more intelligent, layered sub‑queries, meaning it can understand user intent more deeply and surface highly relevant content.
One of the most exciting features is the generative user interface. For example, if you ask a physics question like the three-body problem, Gemini 3 can dynamically generate a simulation. If you’re comparing mortgage options, it might build a custom loan calculator right inside the search result page.
Google also previewed Gemini Agent, powered by Gemini 3, which can do multi-step tasks: think organizing your inbox or planning travel by connecting with Google apps. On the developer front, Google is launching Antigravity, an “agent-first” IDE where AI agents built on Gemini 3 Pro can interact with code editors, terminals and browsers.
This launch aligns with Google’s long-term AI vision. In March 2025, the company had already expanded its AI Overviews to over one billion users and introduced AI Mode for complex queries. Notably, Google is promising that in the coming weeks it will automatically route more complex queries to Gemini 3, while simpler ones will use lighter models, optimizing both power and efficiency.
As Google embeds Gemini 3 into its core Search product and developer tools, it’s signaling that AI is no longer a side project; it’s central to its future. The move also builds on related tools in deep research and AI workflows, like a newly launched research assistant tool.
