One variant will be optimised for edge computing and the other for inference applications, targeted for deployment in Tesla’s Optimus humanoid robots …
http://dlvr.it/TRjrLs