When I started my journey with openAI I overinvested in the “assistants API” and this forum corrected my understanding, pointing out that although lower-level the “chat completions” API was faster – something I then verified for myself.
Since then I’ve circled back over time and used most of the APIs, realtime, assistants (no more), responses, chat completions.
So now before I invest again I wanted to not make the mistake I initially made – and not over-invest — I believe for general purpose the responses API is now the gold standard into the future… is that way off?
Chat Completions is the “gold standard” - drop in the API URL and credentials of any other provider with a compatibility layer, or even their own new useful extensions to include reasoning full-cycle, and you’re off and running with your self-managed conversation.
It seems OpenAI wishes that portability wasn’t there, gating models to only the Responses API, with platform lock-ins such as encrypted content, server-side resources, etc.
I created a topic: feature request-> add Google’s extensions for reasoning to Chat Completions. It seems rather than being heard, it got the “delete” button pressed on it.