Since earlier today, the service is experiencing extreme latency from LLM services as well as transcriber models. Using the same models we've gone from latency of 2,000 ms up to 7,000 ms. please help!

Can you share call IDs, and is this issue still reproduceable?