Rys is a method and set of configurations for fine-tuning and running large language models efficiently on consumer-grade gaming GPUs. It allows researchers and developers to achieve top performance on open LLM leaderboards without requiring expensive hardware.