SESKStuff Everyone Should Know
SESK: Stuff Everyone Should Know · sesk.info/ai-max-tokens-cuts-off

If a reply hits the max_tokens limit, what happens?

It is cut off, maybe mid-sentence or even mid-word.

Start playing free

  1. It is cut off, maybe mid-sentence ✔
  2. The model rewrites it to fit
  3. The API returns an error and no text

Why: max_tokens is a hard stop, not a request to be brief. The model does not plan its answer around it, so if you want short replies, ask for them in the prompt.

Huh, didn't know that: Asking for a number of sentences or paragraphs works better than an exact word count, because models count tokens, not words.

The dad joke
I set max_tokens on my uncle's stories. Now they just stop in the middle of the best p
Like this card?
Play the game

References

Answer: It is cut off, maybe mid-sentence. max_tokens is a hard stop, not a request to be brief. The model does not plan its answer around it, so if you want short replies, ask for them in the prompt.

  1. Claude API Docs: Reducing latency platform.claude.com/docs/en/test-and-evaluate/strengthen-guardrails/reduce-latency
    the response will be cut off, perhaps mid-sentence or mid-word, so this is a blunt technique
    Checked 2026-10-10.

Think this answer is wrong? Tap "Challenge this answer" on the card in the game and tell us why.

Test yourself in the game

Artificial IntelligenceLLM APIs usually charge per request based on what?Artificial IntelligenceHow do AI, machine learning and deep learning fit together?Artificial IntelligenceWhat did Alan Turing call his 1950 test of machine intelligence?

That's stuff everyone should know.