sLLM

1 post working through sLLM adoption decisions in numbers.

• Self-Hosting a 70B Model Is Chartering a Bus - You Need a 25% Load Factor to Beat Taxis

  Self-hosting is a fixed cost and the API is metered, so the deciding number is not monthly tokens but load factor. Add a realistic latency target and a standby node to a long-context workload and the required load factor passes 100%, at which point no amount of volume makes self-hosting cheaper.

  August 23, 2026