Comment by 0xbadcafebee

Came to ask the same. From my really rough understanding, it seems like Unsloth's method allows a slightly higher precision at a higher file size, while PrismML's uses a different approach to achieve a smaller size (and presumably less precision).