Microsoft-Nvidia’s speech recognition model training uses 530 billion parameters for ‘unmatched’ accuracy October 14, 2021 // by Finnovate This content is for members only. Sign up for access to the latest trends and innovations in fintech. View subscription plans. Login