Google has just opened a new chapter in the history of its AI with the introduction of Gemini 3, its most advanced model that inherits and enhances the capabilities of its predecessors. With greater advanced reasoning power, increased intelligence, significant improvements in its multimodal comprehension, and a more robust security layer, Gemini 3 has already been integrated into Google’s extensive ecosystem.
This new technology is already available in the AI Mode for Google AI Pro and Ultra subscribers in the United States. This marks the first time the technology giant has launched a Gemini model in Search from day one. In addition, Gemini 3 has also been rolled out to the Gemini app, the Gemini API in AI Studio, and Vertex AI.
Furthermore, as with its previous versions, the Gemini 3 family comprises several members, with the first being Gemini 3 Pro and Gemini 3 Deep Think (although Google has already announced that more will be introduced in the future).
“Each generation of Gemini has built on the previous one so that you are able to do more. The advances made by Gemini 1 in native multimodality and context window expanded the types and amount of information that could be processed. Gemini 2 laid the groundwork for agentic capabilities and pushed the boundaries of reasoning, enabling assistance with more complex tasks and ideas. This allowed Gemini 2.5 Pro to lead the LMArena ranking for more than six months.
And now, we are introducing Gemini 3, our smartest model, which combines all the features of Gemini so you are able to bring any idea to life,” explained Sundar Pichai, CEO of Google and Alphabet, in the statement issued by the company.
Gemini 3 is the most refined AI model Google has developed to date; its multimodal understanding capabilities (text, images, video, audio, and code), combined with its enhanced advanced reasoning, allow it to analyze, process, and present all types of information to provide you with the best responses in the most appropriate format. “To do so, it combines cutting-edge reasoning, vision, and spatial understanding, along with industry-leading multilingual performance and its context window of 1 million tokens,” states Demis Hassabis, CEO of Google DeepMind, and Koray Kavukcuoglu, CTO of Google DeepMind and chief AI architect at Google.
In addition, Gemini 3 excels at understanding the context and intent behind a query, making it more efficient and delivering more precise and appropriate answers. It has also improved in zero-shot generation and can handle complex requests and instructions. Regarding its personality, it is less flattering (which contrasts with the attitude of the new GPT-5.1),
Another of its strengths is its ability as a model to plan for the long term. This means that “Gemini 3 is better able to help you accomplish things in your daily life. By combining deeper reasoning with improved and more consistent use of tools, Gemini 3 can act on your behalf to carry out more complex, multi-step workflows from start to finish, such as sorting your Gmail inbox or planning a travel itinerary, all under your control and supervision,” asserts Hassabis.
In terms of security, Google affirms that Gemini 3 is its safest model to date, having undergone the most comprehensive security evaluation process of any AI they have developed thus far. Gemini 3 features greater resistance to malicious software and cyberattacks.
The first two members of the Gemini 3 family are the Pro and Deep Think models. The former is already available in a preliminary version, but access to the latter remains limited to Google’s security testers.
Gemini 3 Pro is a multimodal AI with next-generation reasoning that significantly outperforms Gemini 2.5 Pro across key metrics. This AI has the ability to solve complex problems in multiple areas such as science and mathematics. Furthermore, it is capable of providing depth and nuance in its responses, giving you useful and tailored information.
“Its responses are intelligent, concise, and direct, and it prefers to offer valuable information rather than rely on clichés or flattery. It tells you what you need to hear, not just what you want to hear. It serves as a genuine creative companion, offering you new ways to understand information and express yourself, from translating complex scientific concepts by generating code for high-fidelity visualizations to brainstorming creative ideas.”
Gemini 3 Pro is already available in 30 languages (including Spanish, Galician, Catalan, and Basque) within a suite of Google products so that you are able to use it in your everyday life to learn, create, and plan anything. For example, I already have it enabled in my Workspace account.
As its name suggests, this model offers deep thinking capabilities, taking the intelligence, reasoning, and understanding of Gemini 3 even further. Google has explained that it is still testing Gemini 3 Deep Think and evaluating feedback from its testers, but in the coming weeks it intends to make it available to those subscribed to Google AI Ultra.
Google has taken advantage of the announcement of its new AI to introduce Google Antigravity, a development platform based on AI agents, leveraging the power of Gemini 3.
Antigravity transforms AI from a simple tool into an active development partner. Thanks to the advanced reasoning and autonomous programming capabilities of Gemini 3, agents are able to plan, execute, and independently validate complex software tasks.
These agents have direct access to the editor, terminal, and browser, enabling them to work at a higher, task-oriented level. The platform also integrates the Gemini 2.5 and Nano Banana models (for image editing), facilitating end-to-end workflows such as the autonomous programming and validation of a flight tracking application.
Photo: Google
Your email address will not be published. Required fields are marked *
Δ