Cerebras runs Llama 3.1 70B at 2,100 tokens per second, live demo availabletwitter.com6 points·modeless··0 commentsOpen articleSaveView on HN