The AI can consider about 16,384 tokens per request by default (a token is roughly a short word). The dashboard checks the complete request before generating, reserves room for the answer, and blocks requests that do not fit. It never shortens your text silently.
Ways to fit
- Select a shorter passage or fewer PDF pages from your document.
- Use fewer source excerpts.
- Turn off Include the selected course’s settings in Privacy & storage for this request.
- For grading, a long paper may simply be too large for one request.
Last updated October 7, 2026