On X, TiBo posted the idea that somehow caching problems are responsible for the quota burn we all experience. Then another poster appended their statistics after moving to ox-alpha model which is being currrently free on openrouter and statistics is revealing. Massive token flow, almost all terminating in the provider cache.
If OpenAI really mismanaged caching, then this could be the reason behind token burn. Anyone have any similar statistics? Can be this statistic somehow seen on OpenAI subscription page so we can verify it??