PrismML compresses a 27B model to 5.9 gigabytes
PrismML says the new model keeps 98.2% of the initial score, but the company made the tests.
Claimed, not confirmed
PrismML is a start-up that Caltech researchers founded. On September 18, it launched Bonsai 2 27B, a compressed version of the Alibaba Qwen3.8 27B model. The full file is about 54 gigabytes. The new file is below 6 gigabytes. PrismML says the new model keeps 98.2% of the initial benchmark score.
Sources
Posted
