Musk: Grok 4.8 is a model with 2.5 trillion parameters. Using our new C++ software stack training, we will complete the training and launch reinforcement learning this week.

Zhitongcaijing · 2d ago
Musk: Grok 4.8 is a model with 2.5 trillion parameters. Using our new C++ software stack training, we will complete the training and launch reinforcement learning this week.