Hacker News Reader

Looks their performance is better than llama.cpp - https://news.ycombinator.com/item?id=37018989 - and scales to batches of prompts.

Actually no -- that post shows they are not performing measurements and comparisons correctly.

These are not serious people.