i’ve noticed that when i have the same prompt and setting the result of chat and assistant are different from each other.
Not just a little but the whole chat flow is different.
I am wondering what about the cost also since the assistant adds a whole bunch of extra info i dont even want in my prompt.
Escpecially the token count is also completly different than the assistants one
Assistants manages threads for you. with chat completions, you have to manage the threads yourself.
Assistants also gives you access to some rudimentary file search tool, as well as a python runtime for your agent. With completions, you’d have to implement it yourself.
Under the hood, the models should be the same though. So yeah, if you want more control, use completions. If you don’t have much time to slap a PoC together, maybe go with assistants.