## LLama - LLama 3 seems to be on part with GPT-4 on some aspects - 8gb recommended for 16GB M2 - quant version GGUF, llama.cpp, or olalama - https://www.reddit.com/r/LocalLLaMA/comments/1cal17l/llm_comparisontest_llama_3_instruct_70b_8b/ - 70B still best but 8 still not bad ### Setup - Download [[LM Studio]] - Search up `lmstudio llama 3` - Download the one that has the green badge for GPU offloading - Wait for download ## References - [Run Llama 3 locally](https://www.youtube.com/watch?v=KGzF60KERZ4)