09 Model types

Not all LLMs are tuned for the same behavior, and starting with the wrong type is a common reason a capable model feels useless. A base model completes your prompt rather than answering it. Asked “What is the capital of France?”, a base model might continue with “and what is the population of Paris?” instead of answering “Paris.” Base models are useful for pretraining research, fine-tuners, and custom pipelines, and frustrating for everyone else.

The practical split is short. A base model is good for pretraining research, fine-tuningⓘfine-tuningtrainingContinued training of a pretrained base model on a smaller, task-specific dataset to specialize its behavior without retraining from scratch. Chat about this → Open full entry →, and custom pipelines. An instruct model is good for direct instruction following. A chat model is good for multi-turn dialogue with role formatting. A reasoning model is good when the task benefits from extra thinking tokens and verification. A tool-tuned model is good when structured calls, JSON, or function use matter.

For most users, the default starting point is a recent instruct or chat model in a size that fits comfortably in memory. Do not start with a base model unless you know why you want one. What “fits comfortably” means is a memory question, and the self-host track turns it into arithmetic.