Leaked information suggests that Google is building an internal experimental AI model called "Mathematica" based on the DeepThink V3 model. This model is specifically optimized for heavy computing and complex symbolic problem-solving, and is expected to become Google's strongest mathematical reasoning AI model.
Output limit of 65536 token, labeled as "Unstable Experimental Version"
API data shows that the model's internal identifier is "models/deepthink-mathematica-tf-raw-thoughts", supporting a context window of up to 1 million tokens, with an output limit of 65536 tokens. The context window refers to the total amount of text units the model can process at once, while the output limit is the number of text units that can be generated in a single response.
The leaked configuration also includes the label "UNSTABLE_EXPERIMENTAL," indicating that this version is an unstable experimental one. The model also has the "Teamfood" tag - a term used by Google for internal employee testing, meaning that the related service is currently in an internal testing phase.
TestingCatalog notes that Google previously launched the Gemini DeepThink IMO mode for advanced math competitions, and on September 15, released Gemini 3.8 Live and Live Extended Thinking.
