Base Model

The foundation weights before instruction tuning, RLHF, or fine-tuning. The raw next-token predictor.