I've benchmarked its performance at language translation - https://nuenki.app/blog/quasar_alpha_stats

It's very good, top of the leaderboard on almost all languages, and its high score for idiomaticity indicates it's probably quite a large model - and yet it has very fast inference.