Google releases its ‘reasoning’ AI model

Google has introduced a new ‘judgemental’ AI model – but it’s still experimental, and according to our brief tests, there’s certainly room for improvement.

There’s a new model (a mouthful to be sure) called the Gemini 2.0 Flash Thinking Experimental. AI StudioGoogle’s AI prototyping platform. The model card describes it as “best for multimodal understanding, reasoning and coding” with the ability to “reason on the most complex problems” in areas such as programming, mathematics and physics.

a post At X, Logan Kilpatrick, product lead for AI Studio, called the Gemini 2.0 Flash Thinking Experimental “the first step in (Google’s) thinking journey.” Jeff Dean, Principal Scientist at Google DeepMind, Google’s artificial intelligence research unit, he said in his post, he said that the Gemini 2.0 Flash Thinking Experimental is “trained to use thoughts to strengthen their reasoning.”

“We see promising results when we increase the computation of extraction time,” Dean said, referring to the amount of computation used to “run” the model when considering a query.

It’s still an early version, but see how the model handles a challenging puzzle that includes both visual and text clues: (2/3) pic.twitter.com/JltHeK7Fo7

— Logan Kilpatrick (@OfficialLoganK) December 19, 2024

It was recently established by Google Gemini 2.0 Flash model, Gemini 2.0 Flash Thinking Experimental is similar in design to OpenAIs o1 and other so-called reasoning models. Unlike most AI, reasoning models are effectively self-validating it helps avoid some of the pitfalls that normally break AI models.

As a drawback, reasoning models often take longer to obtain solutions—typically seconds and minutes more.

Given the request, the Gemini 2.0 Flash Thinking Experimental pauses before responding, considering a series of related cues and “explaining” its reasoning along the way. After a while, the model summarizes what it considers to be the most accurate answer.

Well – this is what should happen. When I asked the Gemini 2.0 Flash Thinking Experimental how many R’s are in the word “strawberry,” he said “two.”

Your mileage may vary.

in the year After the release of o1has been explosion reasoning models from rival AI labs – not just Google. In early November, DeepSeek, an AI research company funded by quant traders, previewed its first reasoning model, DeepSeek-R1. That same month, Alibaba’s Gwen team opened what it claimed was the first “open” competitor to o1.

Bloomberg informed In October, Google had several teams developing reasoning models. Next to report In November, The Information revealed that the company has at least 200 researchers focused on the technology.

What opened the floodgates of thought pattern? First, the search for new approaches to improve generative AI. Like my colleague Max Zeff recently informed“Brute force” methods of scaling models no longer yield the improvements they once did.

Not everyone is convinced that thinking models are the best way forward. First, they are expensive due to the large computing power required to run them. And while they are carried out good criteria By now, it is very clear that reasoning models will not be able to sustain this rate of progress.