December 20 (Portaltic/EP) –
Google has presented Gemini 2.0 Flash Thinkingits new Artificial Intelligence (AI) model experimental with “reasoning” abilities to answer complex questions about mathematics, programming and physics, which also shows his “thoughts” when resolving user requests.
The tech giant recently introduced Gemini 2.0, the new version of its large language model, which it refers to as the company’s most advanced to power AI agents and that includes improvements in latency and performance.
Now, in order to continue advancing AI capabilities, Google has presented Gemini 2.0 Flash Thinkinga new experimental model of the same family that includes “reasoning” abilities to answer complex questions in subjects such as programming, mathematics and physics.
This has been made known by the Gemini director Jeff Dean in one post on (former Twitter) in which he has detailed that it is a model that is trained to “use thoughts to strengthen your reasoning.”
That is, when responding to requests, Gemini 2.0 Flash Thinking previously shows his “thoughts”, detailing a series of indications with which explain how you are reasoning your answer so that users can understand the process. After that, the model offers the definitive answer to the question in question.
Specifically, this new experimental model is based on the Gemini 2.0 Flash model, also recently announced alongside Gemini 2.0, which guarantees high speed and performance to execute your responses.
Likewise, Dean has announced that, according to the company’s tests, the new model will offer “promising results” when they increase inference calculation timethat is, the amount of calculation used to run the model when answering a question.
According to an example shared by Google product manager Logan Kilpatrick in another post on X, the model is capable of solve a math problem using the data that appears in a photograph and, before answering, it shows all the steps it has taken to reach the solution to the problem in question.
Gemini 2.0 Flash Thinking is still an experimental model, so it has not been officially released. However, it is now available through Google’s AI prototyping platform, AI Studio, for users to test its capabilities.
Add Comment