The battle for AI supremacy is heating up. Almost exactly a week after OpenAI made its , Google today is offering a preview of its next-generation Gemini 2.0 model. In a attributed to Google CEO Sundar Pichai, the company says 2.0 is its most capable model yet, with the algorithm offering native support for image and audio output. “It will enable us to build new AI agents that bring us closer to our vision of a universal assistant,” says Pichai.
Google is doing something different with Gemini 2.0. Rather than starting today’s preview by first offering its most advanced version of the model, Gemini 2.0 Pro, the search giant is instead kicking things off with 2.0 Flash. As of today, the more efficient (and affordable) model is available to all Gemini users. If you want to try it yourself, you can enable Gemini 2.0 from the dropdown menu in the Gemini web client, with availability within the coming soon.
Moving forward, Google says its main focus is adding 2.0’s smarts to Search (no surprise there), beginning with . According to the company, the new model will allow the feature to tackle more complex and involved questions, including ones involving multi-step math and coding problems. At the same time, following a , Google plans to make AI Overviews available in more languages and countries.
Looking forward, Gemini 2.0 is already powering enhancements to some of Google’s more moonshot AI applications, including , the multi-modal AI agent the company previewed at I/O 2024. Thanks to the new model, Google says the latest version of Astra can converse in multiple languages and even switch between them on the fly. It can also “remember” things for longer, offers improved latency, and can access tools like Google Lens and Maps.
As you might expect, Gemini 2.0 Flash offers significantly better performance than its predecessor. For instance, it earned a 63 percent score on HiddenMath, a benchmark that tests the ability of AI models to complete competition-level math problems. By contrast, Gemini 1.5 Flash earned a score of 47.2 percent on that same test. But the more interesting thing here is that the experimental version of Gemini 2.0 even beats in many areas; in fact, according to data Google shared, the only domains where it lags behind are in long-context understanding and automatic speech translation.
It’s for that reason that Google is keeping the older model around, at least for a little while longer. Alongside today’s announcement of Gemini 2.0, the company also debuted Deep Research, a new tool that uses Gemini 1.5 Pro’s long-context capabilities to write comprehensive reports on complicated subjects.
