Comment by draxen

are you evaluating using OpenRouter.ai or other routers/gateways?

they have never been able to keep up with Groq/Cerebras-level inference speeds for some reason.