llm-architectures-explained

/models

Gemini 3 Pro

Google · Gemini · closed weights

Facts and where they come from

Released2025-11labofficial docsmodel card: Model Release November 2025
Licenceproprietarylabofficial docsweights not published
Total parametersnot disclosednot disclosed
Active parametersnot disclosednot disclosed
Context length1,000,000 tokenslabofficial docsmodel card p.1: a token context window of up to 1M
Mixture of expertsyeslabofficial docsmodel card p.2: a sparse mixture-of-experts (MoE) transformer-based model

Architecture

not disclosed No diagram or modelled costs: the model card states the model is a sparse MoE but gives no dimensions.

Sources