Ctrl+Z Therapy
CasesQuick answersGuides10-minute checklistHow we testCartoonSend it in

AI answer in your app cuts off mid-sentence? Check max_tokens

Ship ItChecked against the official source

Source checked 10 Oct 2026 · Published 23 Oct 2026 · Last reviewed 23 Oct 2026 · By Ctrl+Z Therapy (how we test)

The short answer

Every AI API call has an output limit you set in the request. Claude calls it max_tokens: when the answer reaches it, Claude simply stops, and the stop reason is max_tokens. OpenAI's reasoning models return a response with a status of incomplete instead. Claude's docs advise: "Raise max_tokens or continue the response."

What to do

  1. Log why each answer stopped: the stop reason max_tokens on Claude, or a status of incomplete on OpenAI's reasoning models.
  2. If the answer hit the limit, raise the limit or continue the answer instead of showing half of it.
  3. Ask your AI tool: "Find every AI API call in this app. Log why each answer stopped, and if it stopped because of the token limit, raise the limit or offer a 'continue' button instead of showing half an answer."

Good to know

  • On OpenAI's reasoning models, this can happen before any visible text: "you could incur costs for input and reasoning tokens without receiving a visible response."
  • OpenAI recommends reserving at least 25,000 tokens for reasoning and outputs when you start experimenting with those models.
  • For very large limits on Claude, use streaming.

Source

Checked on 10 Oct 2026. Prices, limits and settings change, so check the source before you rely on them.