RT @GoogleResearch: Introducing TurboQuant: Our new compression algorithm that reduces LLM key-value cache memory by at least 6x and delive…