Setting max_tokens / Output Length

The max_tokens parameter caps how many tokens the model can generate in its response.

What does the max_tokens parameter control?