Local Qwen 3.6 35B beats Claude and ChatGPT in enterprise benchmark, prompting migration to on-premise Nvidia hardware · Digg