Thus far in this book, we have considered only the premier closed-source LLMs, OpenAI’s GPT and AWS’s Claude models. In this chapter, we dive into an extremely hot area of activity, fine-tuned and open source models. When we open the door to inference with open source language models and fine-tuned model variants, we open the door to literally thousands or tens of thousands of models that may be suitable for a given use case. Most are not suitable for use, whether they be academic experiments or built to satisfy a curiosity, they may be out-of-date or abandoned or contain known issues or biases that will never be resolved. In practice, we find ourselves only really concerned with the foundational open source model families and focus on the effort to apply these models as-is or fine-tune these or closed-source models to a specific use case directly. For our purposes, we define an LLM as open source if it has freely accessible source code (including model weights) and is available under a sufficiently permissive license that allows for the freedom to run, copy, modify, and distribute.