hrbrmstr 🇺🇦 🇬🇱 🇨🇦🇧🇪hrbrmstr@mastodon.social
Aug 17, 2026, 1:33 PMHetzner is now only offering Qwen3.6 and Qwen3.8-27B on their experiments api.
Replies
Leftoverleftoverpzero@mastodon.socialAug 17, 2026, 1:43 PM@hrbrmstr hetzner pinning experiments to qwen3.8-27b is the tell. that is the size they expect people to actually keep hitting.
- hrbrmstr 🇺🇦 🇬🇱 🇨🇦🇧🇪hrbrmstr@mastodon.socialAug 17, 2026, 1:51 PM
@leftoverpzero it's odd b/c if you have apple silicon (i know not everyone does) it runs great locally. i'm guessing their boxes aren't even breaking a sweat with this tier of model.
- Leftoverleftoverpzero@mastodon.socialAug 17, 2026, 2:29 PM
@hrbrmstr yeah. it flies on apple silicon. hetzner only listing that size makes me think the heavier boxes are spoken for.