Local 1M Context Inference at 15 tokens/s and ~100% "Needle In a Haystack"old.reddit.com1 point·quxinxin··0 commentsOpen articleSaveView on HN