I ran the Qwen3 Coder 30B through LM Studio and with OpenCode(Instead of Claude code). Did decent on M4 Max 32GB. https://www.tommyjepsen.com/blog/run-llm-locally-for-coding