Performance difference between assistant API and chatgpt (based on GPT-4o)

Recently, there seems to be a growing difference in the quality of output between the assistant API and the chatGPT service. Am I the only one who feels that way?
When on earth will the assistant API go GA?