Model qwen3-code, qwen3-ranker, qqen3-embedding deployed on VLLM docker image already available, ollama is easy to start with but too slow. Vscodium+continue or roocode add-on . hardware large vram 24gb more even better