logoalt Hacker News

Iolaumtoday at 7:36 AM0 repliesview on HN

As long as they are transparent about what quant they serve the model and any other optimization they do that also affects performance of inferred tokens.