Consilium: When Multiple LLMs Collaborate

I'm not buying the hype around multi-LLM collaboration. Consilium's paper presents a nice theoretical framework, but it's based on an impractical setup.

Look, running multiple LLMs in parallel is expensive. We're talking about $10,000+ per week for a single instance. That's a hard sell.