B · Normal
On September 14th, Musk said that Grok 4.8 is a model with a parameter scale of 2.5 trillion. It is trained using a new C++ software stack. It will complete the training and start reinforcement learning this week.
Comments