Running Qwen3.6-35B-A3B on 2x 4090 with sglang for GitLab code review
I’m sharing a GitLab review agent I’ve been working on. The idea was to keep code reviews on hardware I run. It’s built on LangGraph.js and uses Qwen3.6-35B-A3B (QuantTrio AWQ, 4-bit) on sglang. The box is two 4090s over PCIe, no NVLink, TP=2.