Google officially announced Gemini 3, the new generation of its big AI model, and said that the word about the biggest jump in capabilities since the first Gemini, with a clear message that it wants to compete equally with OpenAI. The model was announced at an event for investors and partners, and Google describes it as the foundation for all future AI innovations in search, Workspace tools and on Android devices.
See what Google’s Gemini 3.0 will look like:
According to Google, Gemini 3 should better understand longer texts, more complex instructions and a combination of multiple types of content – text, images, video and audio. In addition to the “large” model for servers, there are also smaller, more efficient models intended for mobile phones and laptops, so that part of the AI tasks can be done locally, without constantly sending everything to the cloud. The company emphasizes that the new model is trained to “hallucinate” less, to speak more clearly when it does not know the answer, and to more easily adapt to the rules of the companies that use it.
However, most of the battle is fought in search and office tools. Google plans for Gemini 3 to become the brain behind a new generation of AI answers in search, as well as in gmail, documents and tables, where it can compress long threads, extract tables from text and suggest ready-made answers. Part of the functions will come to Android, for example for smartly summarizing the contents on the screen or leading a conversation with an assistant who “sees” what is on the screen.
The announcement comes at a time when OpenAI is aggressively pushing its models through ChatGPT and various partner products, so Google needs to show that it still has a strong enough response. For our users, this means in practice that in the coming months and years, more and more Google services will receive AI functions based on Gemini – first probably in English and in selected countries, and then gradually in other languages, including Croatian, as the models spread globally.