Thread

1. zkmon · 1791145889 · focus · HN ↗

   I don't get it. It's file size is about 6 times larger than 27B model for the same quant, but the performance improvement is hardly 10% across all benchmarks, according the metrics on it's hf page. Why should one devote so much more hardware for so little benefit?

   1. latentsea · 1791166915 · focus · HN ↗

      Maybe use it and find out. Previously I was only able to run Qwen3.8-27B at acceptable (to me) speeds of 35 t/s on my R9700, but with Strata I'm doing 60 t/s running Qwen3.8-Flash-Next IQ3_XXS. I'm getting better results...

Unofficial Hacker News client; not affiliated with Y Combinator.