[antupis]

Cost is a plus but at least what I see is that getting good response time is even bigger. Something like OpenAI Azure instances are inconsistent and it is far too normal to get a 40sec lag with responses with gpt4-o.