Local embeddings/rerankers more useful than local LLMs for paid users

A Reddit user argues that for those already paying for an LLM service, running local embeddings and rerankers is more useful than running local LLMs. The post suggests focusing on local retrieval and ranking rather than local generation.