2026, Local LLMs

Local models on a Mac Studio

A scratch space for running coding models on my own hardware: Ollama serving Qwen 2.5 Coder, called from a small LangChain client in Python.

Before an AI tool goes into a team's workflow I like to know what it does with nothing between me and the model. This playground runs models locally through Ollama and talks to them with LangChain.

What is in it

  • A LangChain chat client against Qwen 2.5 Coder 14B served by Ollama
  • Notes on setting up the Mac Studio for local inference
  • A folder of ideas and a blog post template

Local LLMs

All work