How to set minimum Tokens for an answer/ output for Assistant API

The title pretty much says it all. I want to set a minimum number of tokens/ words/ characters used for an answer. I have tried the obvious, i.e., telling it in the prompt and the Assistans’ instructions about how long the answer should be in terms of tokens, words, and characters. Thus far, the answers always fall short.

The prompt asks the Assistant to answer four questions with the following character size that should be aimed for.

  1. 1500 characters
  2. 500 characters
  3. 1000 characters
  4. 1000 characters

I have used this article to convert characters into tokens:
https://help.openai.com/en/articles/4936856-what-are-tokens-and-how-to-count-them

One obvious but kind of inelegant solution would be to create four assistants who answer the questions subsequently. However, this would introduce redundancies in the answers as the questions are on the same topic. I could circumvent that somewhat by feeding it the answer of the previous Assistant in the chain, but this seems unnecessarily wasteful of input tokens.

Thanks in advance for your answers!

The AI is very poor at counting characters due to tokenization, and also can’t perceive the tokens themselves.

A better way to define the output is by number of paragraphs, number of sentences per paragraph, number of words per sentence. This allows the response length to be considered in manageable pieces.