Benchmarking Haskell LLM proficiency with AOC puzzles

Thank you! This is really interesting! Can you run any of the models you compare locally? I am interested in how far small models suitable for local execution lack the large ones behind.