Google DeepMind has introduced a new experimental model in its Gemini 2.0 family. Called Gemini 2.0 Flash Thinking Modethis model is designed to showcase the AI’s reasoning process while tackling complex problems, offering a glimpse into how Gemini arrives at its solutions.
“Gemini 2.0 Flash Thinking Mode is an experimental model that’s trained to generate the ‘thinking process’ the model goes through as part of its response. As a result, Thinking Mode is capable of stronger reasoning capabilities in its responses than the base Gemini 2.0 Flash model,” Google says.
The model’s thinking process is returned as the first element of the content.parts list that is created when the model generates the response, the company added
Google CEO Sundar Pichai calls it ‘most thoughtful model’
Noam Shazeer, the AI engineer for whom Google reportedly paid $2.7 billion to rehire from Character.AI, introduced the model in a post on X.
We’ve been *thinking* about how to improve model reasoning and explainability
Introducing Gemini 2.0 Flash Thinking, an experimental model trained to think out loud, leading to stronger reasoning performance.
Excited to get this first model into the hands of developers to try out!
Google CEO Sundar Pichai reshared his post with a caption, “Our most thoughtful model yet:)”
Meanwhile, Jeff Dean, who is the chief scientist at Google DeepMind & Google Research and leads the Gemini team, said that the company is seeing “promising results”.
Introducing Gemini 2.0 Flash Thinking, an experimental model that explicitly shows its thoughts.
Built on 2.0 Flash’s speed and performance, this model is trained to use thoughts to strengthen its reasoning.
And we see promising results when we increase inference time computation!
Limitations of Gemini 2.0 Flash Thinking Mode
Currently available as an experimental model in Google AI Studio and through the Gemini API, Thinking Mode has certain limitations:
Input: Accepts text and images only, with a 32,000 token limit.
Output: Produces text-only output with an 8,000 token limit.
Tool Usage: Lacks built-in tool usage like Google Search or code execution.
