Google Releases Gemini 2.0 Flash Thinking, Surpassing o1-preview Across the Board
Google's Gemini 2.0 Flash Thinking model excels in reasoning and speed, outperforming competitors like o1-preview. It explains its thought process and solves complex problems with ease. Try the experi
A large model that not only can reason but also explicitly shows its "reasoning logic" has emerged.
OpenAI's 12-day continuous release is nearing its end, but its spotlight has clearly been overshadowed by Google's latest offerings. From Gemini 2.0 Flash to Veo 2 and today's Gemini 2.0 Flash Thinking, Google's new releases keep getting more enticing.
As the name suggests, Gemini 2.0 Flash Thinking is based on Gemini 2.0 Flash but has been specially trained to enhance its reasoning ability by using "thoughts."
According to Jeff Dean, Gemini 2.0 Flash Thinking also explicitly displays its thinking process. Below is a demo shared by Jeff Dean on X, where the model solves a physics problem and explains its reasoning process. The entire process took just over a minute.
Although the Flash version is not the full-fledged Gemini 2.0, this Thinking model has already topped the Chatbot Arena leaderboard. Gemini 2.0 Flash Thinking not only ranks first overall but also holds the top position in programming, mathematics, creative writing, and other evaluation tasks (some are tied for first).