A 777M-parameter Mixture-of-Experts model trained from scratch for ~$55. 161M active per token. Base, both chat tunes, and the identity dataset.
-
whoashish115/Moonfrost-777M-Instruct-v2
Text Generation • 0.8B • Updated • 1.85k • 1 -
whoashish115/Moonfrost-777M-Instruct-v1
Text Generation • 0.8B • Updated • 1.83k • 3 -
whoashish115/Moonfrost-777M
Text Generation • 0.8B • Updated • 1.93k • 1 -
whoashish115/Moonfrost-Persona-SFT
Viewer • Updated • 140k • 47 • 1