Google has officially unveiled Gemini 3, its most recent and sophisticated AI model, which the corporation refers to as its “most intelligent” system thus far. The model was integrated into Google products on its launch day, representing the first occasion a new Gemini version is available in Search at the time of release.
Gemini 3 marks what Google executives describe as the next major milestone in their quest for more proficient and context-sensitive AI. In a memorandum announcing the launch, the company stated that the model showcases enhanced reasoning, superior multimodal comprehension, and better awareness of user intent relative to earlier iterations.
What’s different in Gemini 3?
The primary upgrade is reasoning capabilities. Google asserts that Gemini 3 is crafted to understand nuances, tackle intricate issues, and “apprehend depth and context” in ways that its predecessors were unable to. In benchmark assessments, Gemini 3 Pro secured the leading position on the LMArena leaderboard with a 1501 Elo score, outpacing both Grok and Gemini 2.5 Pro.
The model also exhibits considerable progress in academic reasoning. On Humanity’s Last Exam, Gemini 3 Pro scored 37.5 percent without the use of external tools, and on GPQA Diamond, it attained 91.9 percent. It achieved state-of-the-art results across mathematical and multimodal tasks, including video comprehension and factual precision.
A newly enhanced mode known as Gemini 3 Deep Think further amplifies these capabilities. Deep Think boosts scores on nearly every benchmark, including an increase to 41 percent on Humanity’s Last Exam and 45.1 percent on the ARC-AGI-2 challenge. It’s tailored for the most demanding, long-term reasoning tasks.
Where can you access Gemini 3?
Gemini 3 is now operational throughout the Google ecosystem, which includes:
AI Mode in Search, offering more dynamic, generative interfaces and advanced reasoning support. (Currently accessible for Google AI Pro and Ultra subscribers.)
The Gemini app is in the process of being redesigned, including the My Stuff folder and improved shopping functionalities.
Gemini for developers, available through AI Studio, Vertex AI, Gemini CLI, and Google’s new agentic development platform Antigravity.
Gemini Agent, an upgraded system capable of executing multi-step tasks such as inbox organization, appointment scheduling, and more.
How to experience Gemini 3 now
If you wish to try Gemini 3 right away, there are multiple options:
Utilize AI Mode in Search (Google AI Pro or Ultra subscription required).
Launch the Gemini app, where Gemini 3 will be the default model starting today for most global users.
Developers can explore AI Studio or Vertex AI, gaining comprehensive access to Gemini 3 Pro.
Enterprise clients can implement it via Vertex AI and Gemini Enterprise.
Deep Think, the premium version, will be rolled out in the upcoming weeks after further safety assessments.