Google's next generation of Gemini is here to get some work done

A little over a year ago, we entered the era of Gemini. Google’s ever-evolving AI aspirations picked up their latest branding as the company announced its set of Gemini 1.0 models. It only took another month to start moving into Gemini 1.5 territory, and recently we’ve been hearing rumors about when Google might cross the big 2.0 threshold. It looks like those predictions of a December launch were right on the money, as Google introduces Gemini 2.0, designed for the age of agentic AI.

Starting today, you can start trying out the new Gemini 2.0 Flash experimental model. This release delivers not just some big speed improvements over 1.5, but also big upgrades to the model’s ability to generate multimodal output, merging text, audio, and imagery. Devs are getting early access to some of its advanced modes now, but even you can give it a spin by choosing the chat optimized version of Gemini 2.0 Flash with Gemini on the web — and coming soon to the app.

Agentic AI is all about not just thinking, but doing, and empowering these systems to work through more and more complex problems and actually implement solutions. Back at Google I/O in May, we learned about Project Astra, an agent that tapped into our phones’ cameras to function as sort of a next-generation, interactive Google Lens, letting the user ask questions about what they’re shooting.

Google’s still keeping Astra in limited testing, but with the help of Gemini 2.0 it’s already been able to deliver some big upgrades, like support for new languages, improved memory for contextual interactions, and the ability to connect with and take advantage of other Google services like Maps and even Lens iteslf.

Gemini 2.0 is also powering some new efforts, like Project Mariner, a tool that taps into your Chrome browser and is able to interact with websites — this sounds a lot like the rumored Jarvis project we’d heard about before. Google describes this effort as still in the early stages, so it might be a while before you get a chance to try it out yourself. Upcoming Gemini 2.0-powered tools also include Jules for developers, an agent designed to assist with coding and offering GitHub integration.

Of course, half the power of AI models lies in their endless flexibility, and we’re sure we’ll see Google cook up dozens of new ways to tap into Gemini over the course of the next year — Google’s even already thinking about how Gemini can help you have more fun gaming. Sure, AI might one day be coming for all our jobs, but at least it can help us throw together a nice raiding party!