What is an active artificial intelligence model parameter?
Definition
An active artificial intelligence model parameter participates in the computation performed for the current input. Dense models generally use most parameters on every pass, while sparse mixture-of-experts models select only certain expert parameters.
Active parameter count helps describe compute demand but does not fully determine quality or speed. Total model size, memory movement, architecture, quantization, context length and implementation efficiency also matter.
Acronyms and aliases
active AI model parameter variantactive parameter variant
General terms
Related terms
Frequently asked questions
How are active parameters different from total parameters?
Total parameters describe all stored model values, while active parameters are the subset used for a particular computation.
Does fewer active parameters always mean faster inference?
No. Routing, memory access, hardware utilization and software support can limit the expected speed benefit.
Videos explaining active artificial intelligence model parameter
