Google was released a new model of artificiality intelligences he calls a “reasoning” model – but it is currently in the experimental stages. After our short test, there is definitely room for improvement. 
The new model, called the Gemini 2.0 Flash Thinking Experimental, is available in AI Studio, Google's artificial intelligence prototyping platform. The company describes it as “the best model for multimodal understanding, reasoning and codification", with the ability to "calculate the most complex problems" in fields such as programming, mathematics and physics.
It is built on top of Google's recently announced Gemini 2.0 Flash model.
The Gemini 2.0 Flash Thinking Experimental appears to be similar in design to her o1 OpenAI and other models reasoning.
Unlike most AI models, reasoning models are effectively self-policing, which helps them avoid some of the pitfalls that AI models commonly fall into.
On the downside, reasoning models often take longer – typically seconds to minutes longer – to come up with solutions and propositions.
What does it do differently? The Gemini 2.0 Flash Thinking Experimental pauses before answering, considering a series of related questions and "explaining" its reasoning along the way. After a while, the model summarizes what it thinks is the most accurate answer.
