The arrival of Gemini 3 This marks a new step in the race for advanced artificial intelligence. Google presents this model as its most capable system to date, with an emphasis on... complex reasoning, understanding of context, and multimodality (text, image, audio and video), and deploys it directly on products that millions of people use daily.
Beyond the performance charts and technical details, what's relevant for the user is that the experience is quite different from previous versions: More direct answers, less beating around the bush, better structure, and greater depth. when dealing with dense documents, code, educational content, or everyday tasks. And, unlike other launches, Google has chosen to open access to everyone, including those who don't pay for subscriptions.
What exactly is Gemini 3 and how does it improve upon previous versions?
As Google explains, Each generation of Gemini builds upon the previous one.Gemini 1 introduced native multimodality and a long context window, enabling it to handle more types of information and large volumes of data in a single conversation. Gemini 2 focused on agent capabilities, expanding reasoning and planning to address more complex tasks and sustain long workflows.
With Gemini 3The company combines these foundations and takes a leap in several directions at once: more computing power, better reasoning, greater understanding of context, and a deeper integration with the Google ecosystemThe model is designed to capture nuances, both when interpreting a creative idea and when unraveling a technical problem with several overlapping layers.
Demis Hassabis, head of Google DeepMind, defines it as the company's best model in multimodal understanding and programmingWith notable advances in interactive visualizations and reasoning quality, Google executives insist that AI has not yet reached its full potential: improvements in pre-training and post-training, according to the research team, have enabled a leap between Gemini 2.5 and 3.0 as significant as any seen to date.
This pace of progress is reflected in technical benchmarks: in advanced mathematics tests such as AIME 2025, Gemini 3 Pro has a success rate of around 95%.It outperformed models like Claude Sonnet 4.5 and was virtually tied with GPT-5.1 in mathematical reasoning. In broader test batteries, such as ARC AGI 2.0 and Humanity's Last Exam, it also came out on top, suggesting that current AI is already capable of solving problems typically at the doctoral level in science and engineering.
How to activate Gemini 3 Pro and where it is available
Although Gemini 3 Pro is available to everyoneThe change isn't automatic. In the Gemini app and in the search engine's AI Mode, Google keeps the fast mode, based on Gemini 2.5 Flash, selected by default. This means many users don't even realize the new version is available, and if you prefer not to use it, you can Disable Gemini on Android.
To use the new model, open the drop-down menu located below the text box and choose the mode of reasoningwhich appears in many interfaces under the label “Thinking” or “Razoning”. This option is actually Gemini 3 Pro. If you are concerned about privacy, you can Make Google Gemini forget your chat historyOnce activated, the experience changes quite clearly in tasks that require deep analysis or several thought steps.
The rollout is being done in phases, so some users still only see the “2.5 Flash” and “2.5 Pro” options under other brand names. Even so, Google's goal is for the make the Gemini 3 Pro core freely accessiblewhile more specialized features such as Deep Research or future ultra modes will remain tied to paid subscriptions.
This opening contrasts with the strategy followed with the Gemini 2.5 Pro, which was initially reserved for paying customers. Now the company seems to be betting on the mass adoption as a priority, trusting that everyday use will demonstrate whether or not the model is superior to its rivals.
What you'll notice when using Gemini 3 on a daily basis
Beyond performance figures, the differences are clearly noticeable in practice. Users who have compared the quick mode with the new reasoning mode agree that Gemini 3 Pro is faster, more accurate, and sounds less robotic.which is appreciated in realistic tasks such as studying legal regulations, preparing presentations, or writing scripts.
In the analysis of complex documents, for example, when uploading a multi-page Royal Decree and requesting an accessible summary, The mode of reasoning best organizes the informationIt structures sections more thoughtfully and identifies what's truly important. The quick mode works, but tends to be more generic and less clear when the source text is dense and technical.
In the field of writing, the model has also improved the way of expressing oneself: Reduce clichés, avoid unnecessary apologies It uses phrases like “as a language model I can’t…” and cuts out “filler talk.” The result is shorter, more direct, and focused responses, making it easier to integrate into real workflows without so much editing afterward.
A very common use is asking them to adapt texts for different audiences: for example, transforming a technical article into a script for a radio segment aimed at older people with little contact with technology. In this type of assignment, Gemini 3 Pro produces more natural and human-like texts...with less need for subsequent editing to make them sound familiar. The same applies to emails, social media scripts, or step-by-step explanations.
There is also a noticeable leap in the capacity to explain complicated conceptsFrom reviewing how to do square roots by hand to refreshing knowledge of programming or advanced mathematics, the model maintains the essence of traditional explanations but adds better thought-out examples, more visual structures and a better balance between detail and clarity.
Generative interfaces, antigravity, and autonomous agents
One of Google's most striking bets with Gemini 3 are the so-called generative user interfacesIn the search engine's AI Mode, certain queries are no longer answered solely with text, but with interactive visualizations created on the fly: custom maps, simple simulations, or dynamic quizzes that adapt to the user. Tools such as Google Sidekick They show examples of how the company integrates specific assistance in different contexts.
This approach makes AI more than just a chatbot: It acts as a digital agent capable of assembling small applications on the fly.This allows for the creation of learning scenarios or the assistance in making complex decisions without requiring the user to write traditional code. The difference is particularly noticeable in education and planning tasks.
For developers and businesses, Google has introduced AntigravityIt's an agentic development platform that gives AI agents direct access to the code editor, terminal, and browser. This allows them to write programs, run tests, analyze results, and fix bugs in a continuous cycle, with less human oversight.
In technical tests such as WebDev Arena or Terminal-Bench, Gemini 3 stands out for its ability to generate complex web interfacesto manage external tools and coordinate multiple actions simultaneously. The idea is for developers to delegate increasingly larger parts of the software development process to AI, without losing control over key decisions.
This development connects with another growing trend: the emergence of autonomous agents capable of managing digital businesses with almost no human interventionFrom online commerce simulations to experimental platforms where AI manages entire projects, Gemini 3 is positioning itself as one of the driving forces of this new automated economy.
Practical applications: from presentations to everyday purchases
On a more personal level, the new capabilities of the Gemini 3 Pro are noticeable in very everyday uses. One of the most valued features is the creation of presentations based on documents or summariesStarting from a base text, the model can generate slides with a clearer structure (e.g., problem-solution), better readability, and a more logical narrative order than in previous versions.
Reasoning mode also takes a step forward in creative tasks such as design infographics or visual diagramsWhile the quick mode may fall short or return a rather bland result, Gemini 3 Pro is even capable of suggesting Python code to build more advanced visualizations, which is useful for users with a certain technical background.
In daily life, its use is becoming more widespread as shopping assistantA clear example is someone who, needing to furnish a new home in just a few days and on a tight budget, asks Gemini to compare refrigerators, washing machines, and ovens using links to online stores. The model helps interpret confusing terminology, explains each brand's technologies, and prioritizes criteria such as energy efficiency and capacity.
It is also useful for to identify if two products with almost identical names are actually the same model or if they conceal significant differences. By cross-referencing information from various websites, Gemini 3 can clarify whether variations in an appliance's code justify a price jump of more than one hundred euros.
Even in the area of pricing, AI can help: starting from a specific model, it is capable of Visit well-known stores and point out where it's cheapest.It also adds warnings about reputation, after-sales service, or return policies. In some cases, it suggests trying active promotional coupons; they don't always work, but when they do, they can result in significant additional savings.
Gemini 3, images and visual editing
Although Gemini 3 emphasizes reasoning, the visual aspect also gains importance. The model can generate and edit images, making better use of the context that provide user instructions and previous conversation content, even without going through specific "Create Images" menus.
In pure generation, the creations aren't always artistically dazzling, but they do show better alignment with the requirements and greater internal coherence between elements. When it comes to photo editing, the difference is noticeable compared to basic tools: instead of simply blurring an unwanted object, Gemini reconstructs the background by trying to respect patterns, textures, and shapesThis can save images that were previously thought to be lost. You can also compare these capabilities with Google Photos Quick Edit with AI for specific tasks.
Google complements these visual features with options such as Gemini Camera and Gemini LiveThese features allow the model to "see" in real time through the mobile device. This makes it possible to focus on a household problem, an item of clothing, or the ingredients available in the kitchen and receive instant contextual suggestions, as if consulting an expert by looking over your shoulder.
This continued use reinforces the perception that AI is not just a one-off tool, but a constant companion that you turn to while walking, cooking, or travelingOn iPhone, for example, some users describe how they chat with Gemini Live through AirPods with the screen off, combining exercise and learning quite naturally.
Mobile integration, the future of Siri, and its rollout in Europe
One of the most significant movements surrounding Gemini 3 has to do with its possible integration into SiriAccording to leaks reported by international media, Apple and Google are finalizing an agreement whereby the Cupertino firm would pay around $1.000 billion annually to use this model as the "brain" of the new Siri, which is expected to arrive from 2026 onwards.
The approach, however, would follow the Apple style: the 1,2 trillion parameter model would run on Custom servers based on Apple SiliconUnder the company's privacy policies, European user data would not be sent to Google's cloud, but would be processed within Apple's infrastructure, leveraging Gemini's accumulated expertise.
This hybrid architecture seeks to combine Google's computing power with privacy guarantees This is required by European regulations and is used by Apple as a selling point. For iPhone users in Spain and the rest of Europe, it would mean a Siri with conversational, visual comprehension, and reasoning capabilities far superior to the current version, without the need to constantly switch between apps.
Meanwhile, the Gemini app and the search engine's AI Mode are already available in much of Europe, although some more sensitive features—such as certain types of autonomous agents or Deep Think capabilities—may be rolled out more cautiously due to regulatory requirements. The company insists that Gemini 3 is its Safest model to date, with specific improvements to resist command injections, reduce bias, and prevent malicious uses.
Gemini 3 in economics, research, and education
The impact of Gemini 3 transcends individual use. In the professional sphere, the reduction of cost per unit of “intelligence” It is changing the way projects are conceived in sectors such as scientific research, medicine, and engineering. The possibility of delegating entire phases of analysis, simulation, or programming to AI accelerates work cycles that previously took weeks or months.
In education, Google highlights the model's ability to create interactive and personalized learning experiencesFor example, it can translate handwritten family recipes and turn them into a cookbook, transform video lessons into flashcards, or analyze sports recordings to propose training plans adapted to each person's level and goals.
The company has also announced that All university students in the United States will receive a free year of Google AI Prowith access to the most advanced capabilities of Gemini 3 Pro and, in the future, to versions like Deep Think. Such initiatives, if they spread to Europe, could influence how universities integrate AI into their programs.
This rapid progress is not without its drawbacks. Hassabis himself acknowledges that the AI market shows signs of possible bubbleMillion-dollar valuations for startups with little track record, massive investments in data centers, and perhaps an excessive faith in future returns. At the same time, he insists that, even if there were corrections in the sector, Google would continue to rely on AI to improve products like Maps, Gmail, and Search, where it is already seeing clear benefits.
In parallel, concerns about safety and bioethics are growing. The theoretical capacity of advanced models to assist in the design of hazardous biological agents This has prompted defensive initiatives and stricter assessment frameworks, such as those carried out by specialized agencies and the UK's AISI, as well as internal solutions such as the Google's anti-fraud AIThe balance between innovation, privacy, surveillance, and collective security will be one of the major debates in Europe in the coming years.
Competition between giants like Google, OpenAI, xAI, and Anthropic, coupled with the surge of startups launching products almost weekly, ensures that no one can afford to relax. Every advance in benchmarks or practical features forces others to make a move, accelerating a dynamic in which The boundary of what is possible is constantly shifting.
In this context, Gemini 3 is both a cutting-edge piece of technology and a symptom of a deeper shift: artificial intelligence is ceasing to be a curiosity and becoming an everyday infrastructure, integrated into mobile phones, search engines, work tools, and public services. For users in Spain and the rest of Europe, the challenge now lies in learn to take advantage of it criticallyUnderstanding what they can do well, where they continue to fall short, and how it fits into their personal and professional life.
With everything it already offers and what's to come—from Deep Think to its integration into assistants like Siri—Gemini 3 is presented as a model destined to mark this stage of AI: a powerful tool, still imperfect, but increasingly present in how we work, study, stay informed, and make decisions on a daily basis.
