Inference framework Archon promises to make LLMs quicker, without additional costs

Stanford researchers presented Archon, a framework that can cut down on inference costs and allow LLMs to perform better.