standard is the new ultraslow?

source: openrouter.ai where is the speed Tibo promised us? Literally all 5.6 and 6 models are running ultraslow on standard API inference. Funny thing is that subscriptions seems to be served at an even slower pace. My external usage monitor shows that Sol is running at a miserable 26 tok/s as we speak. 🤣 Is this how OpenAI is pacing the frontier?