4 Aug 2026
Hello world!
Welcome to WordPress. This is your first post. Edit or delete it, then start writing!
Welcome to WordPress. This is your first post. Edit or delete it, then start writing!
An average cold start is a number that describes nobody. Ours was 61 ms and the distribution was bimodal.
Streaming looked fine in aggregate because the total time was unchanged. Only the first token moved, and we were not graphing it per region.
Ninety days of notice, a deprecation header on every affected call, and both schemes live at once. It is the only breaking change we shipped this year.
We were counting tokens against total parameters rather than active experts. Everyone affected has been credited; nobody had to ask.
One command, thirty-one regions, and an invoice that matches what you actually served.