It delivers ~450,000 tokens/second on Apple M1/16GB

Show HN: WinkNLP introduces key sentence extraction
github.com