Polymarket, citing Bloomberg, says startups are increasingly building custom AI models with open-weight alternatives to cut costs and reduce dependence on OpenAI and Anthropic. A commenter sharing the report argues that a compute capacity crunch makes cheaper, higher-quality specialized models especially valuable. They outline two ways to cut inference costs—the cost of running models: optimize inference, which they say is getting harder, or train models to use fewer tokens and improve smaller models at specific tasks, aiming for lower costs at the same performance.