Assistant GPT 3.5 model API poor performance


We are using GPT 3.5 turbo 26k model assistant, and trigger thread creation, runs and message creation from the API. The performance is bad, leading to the application being unpresentable, thus useless.
We understand that GPT4 turbo, the latest models etc., are much slower, but shouldnt the most decent model (regarding reasoning) of the past releases be more decent also in response times?
Or is this because of the beta version?

The average response time is 10-15sec, sometimes even more than that.

Please, we would like to have an official answer as this is a company need, not just experimenting.

Thank you.

Hi and welcome to the Developer Forum!

The Assistants API is still in development and is not suitable for a production environment. You should always make sure your clients and customers are aware of the cutting edge and rapid development and related issues with AI products.

All API based services will encounter outages and communication issues, AI is especially prone to these issues as so much of what is being offered is totally new and there are no previous systems to look back on for methodologies.