Search
Google is in the process of developing an experimental AI model, internally codenamed Mathematica, which leverages the foundation of the DeepThink V3 model. This particular model has been fine-tuned to excel in managing intensive computations and resolving intricate symbolic mathematical problems. It is anticipated to emerge as Google's most formidable AI model for mathematical reasoning. Internally, it is designated as 'models/deepthink-mathematica-tf-raw-thoughts', boasting a context window capable of accommodating 1 million tokens and an output capacity of up to 65,536 tokens. Presently, the model is undergoing internal testing and is regarded as an experimental version with inherent instability. Prior to this, Google has already rolled out the Gemini DeepThink IMO mode and subsequently introduced the Gemini 3.8 Live and Live Extended Thinking features on September 15.
-
C114 Communication Network -
Communication Home
