Q4_K_M is bigger than Q4_K_XL

#7
by ru5h - opened

Something is wrong :

Q4_K_M
138 GB

Q4_K_XL
131 GB

Unsloth AI org

This is completely normal. Our dynamic quants are actually usually smaller

Something is wrong :

Q4_K_M
138 GB

Q4_K_XL
131 GB

XL is the suffix for unsloth's UD quants and should not be directly compared to llama.cpp standards in file size

ru5h changed discussion status to closed

Sign up or log in to comment