The pace of large-scale model releases is accelerating

In 2017, only two models exceeded 10²³ FLOP in training compute. By 2020, this grew to four models; by 2022, there were 32, and by 2024, there were 174 models known to exceed 10²³ FLOP in our database, and 99 more with unconfirmed training compute that likely exceed 10²³ FLOP. As AI investment increases and training hardware becomes more cost-effective, models at this scale come within reach of more and more developers.