base model

A model that has only been pretrained -- it completes text plausibly but has no built-in tendency to follow instructions or hold a conversation.

A base model is a language model immediately after pretraining, before any instruction tuning or further post-training. It has learned to predict plausible continuations of text, but has no special tendency to answer a question directly, follow an instruction, or behave like a conversational assistant -- those behaviors come from post-training (chapter 19), not pretraining alone.

Examples

Asked "What's the capital of France?", a base model might just continue with "and what's the capital of Germany?" instead of answering, since it only learned to continue plausible text.