4 ms·

Last time I tried this service they served me lobotomized models with horrible latency and high error rate