New Google AI model now easier to use

Google has just released an improved version of its AI model called Gemini 2.0 Flash Thinking Experimental. This new version, which you can try out for free in Google AI Studio or through the API, was first hinted at back in December.

This updated model is designed to think more like a human, showing its thought process as it solves problems. It can handle much bigger tasks, with a new feature allowing it to work with a million tokens of context, which is great for dealing with large codes or lots of documents.

It also supports running code directly, which means it can use tools better. This version produces longer responses and makes fewer mistakes compared to earlier models. It’s shown significant improvements in math and science, scoring 73.3% on the AIME2024 math test and 74.2% on the GPQA Diamond science test.

Demis Hassabis from DeepMind pointed out how quickly this model has advanced since its initial release, building on years of research started with projects like AlphaGo. Sundar Pichai from Google added that they plan to keep enhancing this model, promising even more exciting developments in the future.