GPT-4o-mini randomly much slower than GPT-3.5-turbo

The service “is what it is”. OpenAI has previously also penalized low payment tier users with lower output rates.

Perhaps the gpt-3.5-turbo models are underutilized, just waiting for only you?