Google's TurboQuant AI-compression algorithm can reduce LLM memory usage by 6x

TurboQuant makes AI models more efficient but doesn't reduce output quality like other methods.