Google has introduced a new experimental AI model, Gemini 2.0 Flash Thinking Experimental, which the company claims is capable of "reasoning over the most complex problems" in fields such as programming, math, and physics. The model is available in AI Studio, Google's AI prototyping platform, and is described as "best for multimodal understanding, reasoning, and coding."

According to Logan Kilpatrick, who leads product for AI Studio, Gemini 2.0 Flash Thinking Experimental is "the first step in [Google's] reasoning journey." Jeff Dean, chief scientist for Google DeepMind, Google's AI research division, explained that the model is "trained to use thoughts to strengthen its reasoning." Dean noted that the model shows promising results when increasing inference time computation, referring to the amount of computing used to "run" the model as it considers a question.

Gemini 2.0 Flash Thinking Experimental is built on Google's recently announced Gemini 2.0 Flash model and appears to be similar in design to OpenAI's o1 and other so-called reasoning models. Unlike most AI, reasoning models effectively fact-check themselves, which helps them avoid some of the pitfalls that normally trip up models. However, a drawback of reasoning models is that they often take longer – usually seconds to minutes longer – to arrive at solutions.

In practice, Gemini 2.0 Flash Thinking Experimental pauses for a matter of seconds before responding to a prompt, considering a number of related prompts and "explaining" its thinking along the way. After a while, the model summarizes what appears to be the best answer. However, in testing, the model struggled with simple tasks, such as counting the number of R's in the word "strawberry," incorrectly responding with "two."

The release of Gemini 2.0 Flash Thinking Experimental comes amidst an explosion of reasoning models from rival AI labs. In early November, DeepSeek, an AI research company funded by quant traders, launched a preview of its first reasoning model, DeepSeek-R1. That same month, Alibaba's Qwen team unveiled what it claimed was the first "open" challenger to o1. The surge in reasoning models is driven by the search for novel approaches to refine generative AI, as traditional "brute force" techniques to scale up models are no longer yielding the improvements they once did.

The development of reasoning models like Gemini 2.0 Flash Thinking Experimental has significant implications for the future of AI research and development. As AI models become increasingly complex and sophisticated, the ability to reason and fact-check themselves will be crucial in avoiding pitfalls and ensuring accurate results. While Gemini 2.0 Flash Thinking Experimental is still in its experimental stages, it marks an important step in Google's "reasoning journey" and has the potential to pave the way for further advancements in AI research.