Training computation vs. parameters in notable AI systems, by domain

Computation is measured in total petaFLOP, which is 10¹⁵ floating-point operations estimated from AI literature,albeit with some uncertainty. Parameters are variables in an AI system whose values are adjusted during training toestablish how input data gets transformed into the desired output.

Number of parameters101001,00010,000100,0001 million10 million100 million1 billion10 billion100 billionTraining computation (petaFLOP)<0.1BiologyGamesImage generationLanguageMultiple domainsOtherSpeechVisionJul2,1950Feb9,2020

Select systems

Sort by
Name
    All systems
    Training computation vs. parameters in notable AI systems, by domain

    Interactive visualization requires JavaScript