Hey all, This might be a great article for infrastructure teams managing LLM costs at scale. The provider evaluation section is quite useful: same model, different providers, meaningfully different cache hit rates and eval scores. ito.ai/blog/measuring-100b-tokens-week-open-weight-models