Liked Muse, so I cut the 30B model in half by width, distilled it back, and it does 57 of 60 tool tasks its parent does 60 of

Excerpt

I've liked how Muse-Glimmer worked, so I wanted to see if I could produce a smaller "kid" out of it. Ornith's sharp decisions on when to think and which tool to call were the other thing I liked, so Ornith-1.0-9B got to be the policy teacher while the parent wrote the words. No RL anywhere, distillation only. I present to you Xyntetik-Kvist-14B . What it is good for Smaller than the Muse parent but still manages most tool tasks: 57 of 60 held-out closed-loop tasks (contacts, weather, flights, cu