Researchers at MIPT have introduced the HARP algorithm, which enhances the performance of highly compressed large language models by reducing errors by up to 15%.
The Linux kernel is transitioning to a modern upstream version of the LZ4 compression algorithm, promising performance gains of up to 21% and simplified maintenance.
PrismML has launched the Bonsai 2 27B language model, which is 9–10 times smaller than the original Qwen3.8 27B while retaining 98% of its benchmark performance.
PrismML has raised $22.25 million to develop compression technology that enables large language models to run locally on PCs and smartphones.
