Last time, I launched an OpenAI-compatible API using the Qwen3.8-Flash-Next NVFP4 configuration. (Details below) Since I got ...