cc-coder-GGUF

cc-coder GGUF — CC: Tweaked Lua coding model for Ollama / LM Studio

Quantized q4_k_m build (~1.9GB, runs comfortably on any modern machine — CPU-only works too) of minecartchris/cc-coder, a Qwen2.5-Coder-3B fine-tune for writing CC: Tweaked (ComputerCraft) Lua: turtles, peripherals, rednet, monitors, and community libraries.

Evaluated by actually running its generated programs in a CraftOS-PC emulator: 74% of everyday CC task generations run cleanly, 97% parse. Full methodology on the adapter page.

Run with Ollama

Straight from this repo, no download step:

ollama run hf.co/minecartchris/cc-coder-GGUF

Then ask away:

>>> Write a turtle program that strip-mines a 2x1 tunnel for 32 blocks,
    placing torches every 8 blocks, and returns home when done.

Run with LM Studio

Search for cc-coder in LM Studio's model search (or paste minecartchris/cc-coder-GGUF), download the q4_k_m file, and chat. Suggested settings: temperature 0.4, context 4096.

Notes

• The model prefers built-in CC APIs (turtle, rednet, peripheral, fs, os.pullEvent, textutils) and knows the correct require idioms for popular libraries (Basalt, Pine3D, PixelUI, ccryptolib, ...)
• 3B model: always eyeball the logic before trusting a turtle with your diamonds
• Qwen Research license inherited from the base model (non-commercial restrictions apply)