Outputs that are too long waste readers' time; too short and they omit what matters. Length is easy to control once you ask precisely.
Be Specific
- "In about 100 words."
- "Three bullet points, each under 15 words."
- "One paragraph, then a table."
- "A single sentence."
Word and sentence counts are followed approximately; structural constraints (number of bullets or sections) are followed more reliably.
Explain the Reason
"Keep it under 80 words because it will be read on a phone notification" helps the model prioritise.
Prioritise Content
Tell the model what to keep if it must cut: "If space is tight, keep the deadline and cost; drop the background."
Avoid Padding
Ask for no preamble, no restating the question and no closing summary when you don't need them.
Maximum Token Settings
API settings that cap output tokens stop generation abruptly, which can cut off mid-sentence. Use them as a safety limit, and control length through instructions.
Long Outputs
For long documents, generate section by section from an outline, rather than asking for everything at once.
Check With Examples
Show an example of the desired length and density; models imitate examples closely.