Google begins testing Gemini 2.0 Experimental Advanced for subscribers

Google is now letting Gemini Advanced customers test a new 2.0 Experimental Advanced model, after starting Gemini 2.0 Flash Experimental in the previous week.

In the model picker on desktop and mobile web (it's not on the app just yet), the new test appears as a "2.0 Experimental Advanced". If you decide to test it out, you'll be using "Gemini-Exp-1206". According to Google, the model offers significantly improved performance on certain tasks.

The tasks where this model will reportedly excel include coding, math, reasoning, and instruction following, which are quite complex tasks. According to the Mountain View tech giant, the model will handle such complex tasks with more ease.

The generative AI model should be better at solving math problems and providing multi-step instructions to craft a tailored business plan.

Right now, this test is in an "early preview", so it may experience some issues and might not work as polished. In fact, the model does not have access to real-time information and is not currently compatible with some of the existing Gemini features while it's in its experimental state.

So far, Gemini Advanced has granted subscribers access to several interesting generative AI models. Those include the 1.0 Ultra and the 1.5 Pro models. It's possible that this one in particular is 2.0 Pro or even higher.

According to Google, you can expect more Gemini 2.0 model sizes in January, which is when 2.0 Flash will become available for developers.

In order for you to get access to such tests, you need to have a Gemini Advanced subscription. The subscription costs $19.99 per month with Google One AI Premium.

I really like generative AI and I always get hyped when a new model is in testing. I feel like the tech is still far from reaching its full potential and I'm more than curious to see where it will lead when it does.