Pollux – a natively vector quantized LLM with 0.76 bits per parameter

Here's a brief summary of the Pollux project: Pollux is an open-source, natively vector quantized Large Language Model (LLM) that achieves a remarkable 0.76 bit