Very advanced reasoning models trained on many different datasets, such as qwq and sonnet 3.5, these models excel in math, RP, explaining things.