A base model is the raw pretrained text-completion model; an instruct (or chat) model has been further tuned to follow instructions and hold a conversation.
A base (or "foundation") model is trained to predict the next token over huge amounts of text. It's powerful but only knows how to continue text — ask it a question and it may just keep writing more questions. Base models are the starting point for further training and for some completion or fine-tuning use cases.
An instruct/chat model takes that base and fine-tunes it (often with instruction data and preference tuning like RLHF/DPO) so it answers questions, follows directions, and respects a chat format with roles. This is the version almost everyone wants for assistant-style use.
When downloading, check the suffix: "-Instruct" or "-Chat" means it's ready to talk to; the plain/base version expects you to either fine-tune it or use raw completions. Using a base model where you wanted a chat model is a common beginner surprise.
Fine-tuning · LoRA · Inference · Parameters (the "B" / billions)
All explainers → · Browse models →
Open the free Spanvero advisor → · Honest, $0-markup. © 2026 Cynosure LLC.