Full threadsp332·Still Llama-based.Anyway, I'm glad they went with 30B over 13. It's more flexible with requests about how output is formatted.View on HN