Latest news from Google I/O 2025

The wait is over for one of the most important developer events of the year: Google I/O. The technology giant is immersed in one of its benchmark conferences, where they are launching numerous news about the latest tools and developments they are working on. A long list of interesting advances in generative AI that are going to completely change the industry.

We didn’t want you to miss anything, so here’s a summary of the most important news!

Google I/O 2025 Recap

Below, we break down the most important releases and news, but we start with a 10-minute summary of the highlights:

Gemini 2.5

Gemini has many new features such as Google AI Studio updates, native audio output and live API, native audio dialog, asynchronous function calling, computer usage API, URL context, model context protocol, and more…

Google Beam

It is an AI-based video communications platform. Beam uses a new state-of-the-art video model to transform 2D video streams into a realistic 3D experience.

To do this, it uses an array of six cameras and AI to combine video streams and render them to you in a 3D light field display. It has near-perfect head tracking, down to the millimeter, and at 60 frames per second, all in real time. The result is a much more natural and immersive conversational experience. In collaboration with HP, the first Google Beam devices will be available to early adopters later this year.

Project Mariner

It is an early research prototype developed at Gemini 2.0 that explores the future of human-agent interaction, starting with the browser. It is able to understand and reason with information on the browser screen, including pixels and web elements such as text, code, images and forms. And then it uses that information through an experimental Chrome extension to perform tasks for you.

Agent Mode

They have also started adding agent features to Chrome, Search, and the Gemini app.

For example, the new Agent Mode in the Gemini app will help you do even more things. If you’re looking for an apartment, it will help you find listings that match your criteria on sites like Zillow, adjust filters, and use CCM to access listings and even arrange a viewing.

Coming soon, subscribers will be able to access an experimental version of Agent Mode in the Gemini app.

Smart personalized answers in Gmail

Google is working on the best way to bring research into the real world. So, after giving permissions, Gemini can use your relevant personal context in your Google apps in a private, transparent way that is totally under your control.

One example is Personalized Smart Replies in Gmail, which will emulate your regular greeting and capture your tone, style, and even your favorite words to generate a more relevant response that sounds like you wrote it. Personalized Smart Replies will be available to subscribers by the end of the year.

Gemini 2.5 Flash

It introduced a new version of its leading Gemini 2.5 Flash model, which will feature increased performance in complex coding and reasoning tasks, optimized for speed and efficiency.

It also features greater transparency and control, as thinking summaries are now available in its 2.5 models, with the goal of helping developers further manage costs and control how models think before responding.

Both Gemini 2.5 Flash and 2.5 Pro will appear in Google AI Studio and Vertex AI in preview, with general availability of the former in June and the latter coming soon.

Gemma 3n

It is their latest open, fast, and efficient multimodal model, designed to run seamlessly on phones, laptops, and tablets. It supports audio, text, image, and video, and is now available for preview in Google AI Studio and with Google AI Edge.

Also unveiled was MedGemma, a ready-made model for multimodal understanding of medical text and images, designed for developers to tailor and build their healthcare apps. Or SignGemma, an upcoming model that translates sign languages to spoken text.

Gemini Diffusion

This is a new text model that is touted as “not just fast, but lightning fast”.

The demo launched during the event showed that it can generate text at five times the speed of its fastest model to date, while matching its encoding performance.

Project Astra

This project explores the possibilities of a universal AI assistant capable of understanding the world around us.

Its main features are: better dialog, new tool usage, better memory, and improved latency.

You can see an example in this video:

Google Search

Gemini models are also helping to make Google search more intelligent, proactive, and personalized.

Since launching last year, View Created with AI has reached more than 1.5 billion users and is now available in 200 countries and territories. They have observed that users using View Created with AI are more satisfied with results and do more searches.

In fact, in their largest markets, such as the U.S. and India, AI-powered Views are driving more than 10% growth in the types of searches they are showing.

AI Mode

Closely related to the previous point, for those who want an all-AI search experience, they have introduced the new AI Mode. It is a total reinvention of Search.

Thanks to more advanced reasoning, you can ask longer and more complex questions in AI Mode. In fact, early users who have tested it have made queries that are two to three times longer than traditional searches, and you can go even further with follow-up questions.

All of this is available in a new tab in Search.

Gemini Live

As for apps and Gemini Live, this is a feature that is being very well received by users, and they are going to make the camera and screen sharing features available to everyone, including iOS users, and will soon connect them with your favorite Google apps to make support more seamless.

Veo 3 and Imagen 4

It’s been a big day for video models, as Veo 3 has been unveiled as their most advanced video model to date, which now includes native audio generation.

But it’s not the only one, as we’ve also seen the latest features of Imagen 4, their most efficient and capable image generation model. Both are now available from Gemini.

Flow

It’s a new AI filmmaking tool designed with creatives for the next wave of storytelling.

As shared by Google, it is the only cinema AI tool custom-designed for Google’s most advanced models – Veo, Imagen, and Gemini. Flow can help storytellers explore their ideas without limits and create cinematic clips and scenes for their stories.

It’s still in its infancy, but it could be the future of something much bigger for the audiovisual industry.

Android XR

Google’s second era of smart glasses is off to a spectacular start. Google has announced its collaboration with Samsung, Gentle Monster, and Warby Parker to create smart glasses that people will actually want to wear.

Warby Parker is well known as a direct-to-consumer eyewear brand that makes it easy to get fashionable glasses at a relatively affordable price. Meanwhile, Gentle Monster is currently one of the most popular eyewear brands.

Another aspect to note is that Google also seems to be betting on Samsung for the hardware of its XR glasses, which will be manufactured by Xreal under the name Project Aura.

As for the functions of these XR glasses, Google emphasized that they are an excellent tool to use with Gemini.

And here is the summary of Google I/O 2025. Don’t miss the latest news in our networks and newsletter!