Right now, the cost to run model for inference in GPU is cost-prohibitive for most ideas, projects, and bootstrapping startups compared to just using chatgpt API. Once you are locked in the ecosystem ...