Qwen3-Coder

Best Local LLM for Agentic Coding on a Single RTX 4090

A 4090 owner benchmarks GLM-4.7, Nemotron-30B, and Qwen3-Coder for local agentic coding via llama.cpp.