1M Tokens/s: Scaling Qwen 3.5 27B on 96 B200 GPUs with vLLMmedium.com3 points·m4r1k··0 commentsOpen articleSaveView on HN