Four open-source projects, Unsloth, Axolotl, TRL, and LLaMA-Factory, are currently prominent in large language model fine-tuning. These projects are built on a shared foundation of PyTorch and Hugging Face, but differ in their approach to optimization. Unsloth focuses on rewriting kernels, Axolotl composes parallelism strategies, TRL defines trainer APIs, and LLaMA-Factory prioritizes model coverage. By comparing these frameworks, developers can better understand the trade-offs involved in fine-tuning large language models. This comparison matters as it can help inform decisions on which framework to use for specific projects.