Foundation vs. Instruct vs. Chat Models: One Question, Three Answers
A tutorial demonstrates the distinctions between foundation, instruct, and chat models in large language models. It uses the SmolLM2-135M family, runnable on Google Colab without a GPU, to illustrate how models evolve from basic text completion to helpful assistants. The tutorial highlights that foundation models predict the next token, instruct models are fine-tuned on instruction-response pairs and require specific formatting, and chat models build upon instruct models by adding conversational memory. AI
IMPACT Clarifies the distinct capabilities and training stages of LLMs, helping users understand model behavior and limitations.