Imagine spending $5k to run a 32B param llm locally.. You could run much more capable open source models through Openrouter for years running 24/7 at 50tps. Thi

Imagine spending $5k to run a 32B param llm locally.. You could run much more capable open source models through Openrouter for years running 24/7 at 50tps. This will never make sense to me.