Google's AI already works in Catalan (and Siri will eventually do so)
The internet giant is officially supporting our language with the Gemini 3 Pro, which will also bring Catalan to Samsung devices and, by extension, to Siri on iPhones. All this while Sundar Pichai warns that the AI party won't last forever.
BarcelonaIf anyone needed proof that artificial intelligence has ceased to be a toy and has become critical infrastructure, they only had to look to California last Tuesday. Google unveiled Gemini 3 Pro, its new beast of logic and multimodal processing. But for users here, the news isn't the benchmarks Not even quantum reasoning ability, but a simple line in the release notes that puts an end to a long-standing anomaly: Catalan is now, officially, a compatible language.
Until now, our relationship with Google's AI was one of mutual tolerance. The chatbot understood us because it had devoured the entire internet, Wikipedia included, but there was no guarantee of service. How are we doing? explain to the ARAThis precariousness led to grotesque situations like that of Meta's AI on WhatsApp, which, despite knowing Catalan perfectly, had—and still has—a corporate order to play dumb and respond with a laconic "I don't speak your language" when addressed in the language of Quim Monzó. With Gemini 3 Pro, Google emerges from its linguistic ambiguity and places Catalan at the same functional level as ChatGPT (OpenAI), Microsoft Copilot, and, surprisingly, Alibaba's Qwen, thus expanding the circle of major platforms.
The list of thirty languages and the geopolitics of language
Catalan isn't the only language joining the party. With Gemini 3 Pro, Google has added formal support for thirty new languages which, if ordered by number of speakers, clearly reveal where Google is looking for its next billion users. The absolute priority is the Indian subcontinent and Southeast Asia. Leading the list are demographic giants like Bengali—with over 230 million speakers—and Urdu, closely followed by Indonesian, Vietnamese, Telugu, Marathi, and Tamil. Key languages for commercial strategy, such as Punjabi, Gujarati, Thai, and Persian, are also included. Further down the list are Burmese, Amharic (Ethiopia), Ukrainian—for obvious geopolitical reasons—and Malay.
Catalan appears in the European blog on "advanced user retention," alongside our Basque and Galician neighbors, and there are also other mid-sized state languages such as Croatian, Serbian, Slovak, and Lithuanian. The fact that Google puts us in the same technical category as languages with their own state and tens of millions of speakers, like Vietnamese or Turkish, demonstrates that digital pressure and the quality of the Catalan data corpus pay off.
Thinking, seeing, and programming interfaces
Beyond language, what does Gemini 3 Pro actually bring to the table? The list of new features is technically dense. The biggest innovation is the "Deep Thinking" mode. Until now, chatbots have been like that student who raises their hand before the teacher finishes the question: fast, but often inaccurate. Gemini 3 Pro incorporates a reflection phase in which the model evaluates several logical paths before starting to respond. This translates into a significant improvement in math: in the MathArena Apex test, Gemini 3 Pro achieves a 23.4% success rate, sweeping aside the paltry 1.0% of GPT-5.1.3. It's no longer just a stochastic parrot; it's now starting to become a reasonable calculator.
The second new feature is generative interfaces. If you ask it to plan a weekend route through La Garrotxa, it won't just return text: it can generate, in real time, an interactive map and hotel listings, and create a short-lived mini-application just for you. This is combined with Vibe Coding, a programming method already offered by other AIs like Claude, in which the user describes the vibe –the feeling– that you want for an app, and the AI takes care of writing the code, designing the visual elements, and deploying it.
Google's AI in mobile
The interesting stuff isn't happening in the web browser, but in our pockets. Gemini is becoming the unsung AI of Android phones. Samsung's Galaxy AI already integrates features that draw directly from Google's technology. The most striking example is the integration of the "Nano Banana" image model (officially Gemini 2.5 Flash Image), which has been embedded in the image gallery, allowing you to transform a selfie into a comic book character or erase elements from a photo without having to open any external apps. Samsung has been clever: instead of creating its own AI from scratch, it has "tuned" Google's to offer a better experience. When Galaxy AI is updated in Gemini 3, Catalan should also arrive.
And the same should happen with Chinese manufacturers. Xiaomi's HyperOS 2.0 integrates Gemini under the HyperAI brand to connect applications, allowing a calendar entry to automatically generate a smart reminder. Oppo, for its part, offers Mind Space, in which Gemini organizes the user's chaotic thoughts and files. However, the most fascinating twist affects iPhone users. According to Bloomberg, Apple has agreed to pay Google around $1 billion annually for Gemini to be the brain of the next generation of Siri, expected in the spring of 2026. This could benefit Catalan speakers: Siri, which has been avoiding Catalan for years, could suddenly inherit Google's new linguistic capability. After so many years of grumbling for Apple to add our language to Siri, it turns out the solution will come from the competition. If you have an iPhone, you may end up talking to Google, even if the voice is the same old Siri. How much does AI cost?
Individual use of Gemini 3 Pro is free with the mobile app; on a computer, a monthly subscription of around €22 is required, which includes storage and other features. For businesses and developers—the market that AI companies are truly interested in—a price war is already underway: Google has entered the fray and offered a price aimed at undermining OpenAI, although it's not the cheapest. Gemini 3 Pro sells for €1.90 per million tokens The Gemini 3 Pro—the unit of measurement for AI—starts at €11.40, more than OpenAI's GPT-5.1 at €1.19/€9.50, but considerably less than Anthropic's Claude Sonnet 4.5 at €2.85/€14.25. Google offers a broader contextual window that justifies the higher price compared to OpenAI for complex tasks, such as analyzing entire books at once. Meanwhile, in India, the operator Jio is already giving away Gemini 3 Pro to its customers, reminding us that in the AI economy, user data from emerging countries is worth as much as, or even more than, the euros of Europeans. All this technological deployment comes, paradoxically, at a time of maximum uncertainty. Sundar Pichai, Google's CEO, took advantage of the launch to drop a bombshell on the BBC: "When the AI bubble bursts, no company will be immune, not even Google." It's unusual to hear an executive talk about "irrationality" and a "bubble" while presenting their flagship product. Pichai admits, somewhat reluctantly, that the current pace of investment—in which... They burn billions in chips– it's not sustainable. Google has the money to survive the coming winter, but many of the start-ups that are flourishing around AI today, perhaps not.
For the Catalan-speaking user, the lesson is twofold. We celebrate that our language now has an official passport to the AI elite, but let's do so with the skepticism of someone who knows they are using a technology subsidized by an investment frenzy that, sooner or later, will have to be corrected. Let's take advantage of it while it lasts.