Google on Thursday
released Gemini 2.0 Flash Thinking Experimental, an AI model that uses runtime "reasoning" techniques similar to OpenAI's o1 to achieve "deeper thinking"
for queries.
The announcement, made on X by Google CEO Sundar Pichai, explained it all. “Our most
thoughtful model yet," he wrote.
The company did not mention a use for advertising, but it is still considered an experimental model.
Noam Shazeer, vice president of engineering, Gemini co-lead at Google DeepMind, added that Google had been working to improve AI reasoning for "some
time," and that this is an "experimental model trained to think out loud, leading to stronger reasoning performance."
advertisement
advertisement
On the developer documentation page, Google explains how Thinking Mode is capable of stronger reasoning capabilities in its responses than
the base Gemini 2.0 Flash model, previously Google’s strongest model. He posted a
copy of the model to demonstrate how it works.
The company says the algorithm can tackle problems across fields such as programming, physics and math.
In
one demonstration video, Google showed Gemini 2.0 Flash Thinking Experimental tackling a logic puzzle, which required
the model to analyze a photo of four billiard balls.
It developed a correct answer after determining that the photo needed to be flipped. The model developed the solution through a
process in which it tried several different approaches one after another.
Thinking Mode is available as an experimental model in Google AI Studio, and for direct use in the Gemini API.
The way the model's thoughts are
returned depends on whether the user uses the Gemini API directly or makes a request through Google AI Studio.