{"version":"2.1.290","anchor":"max-output-token-continuation-prompt-tells-the-model-to-rest","canonical_anchor":"max-output-token-continuation-prompt-tells-the-model-to-rest","heading":"New instructions for continuing a reply that was cut off","tier":"notice","area":"System Reminders","scope":"individual","heads_up":false,"url":"https:\/\/changelogs.core-directive.com\/v\/2.1.290\/e\/max-output-token-continuation-prompt-tells-the-model-to-rest","release_url":"https:\/\/changelogs.core-directive.com\/v\/2.1.290","markdown":"### New instructions for continuing a reply that was cut off\n\nA new message for cut-off replies tells Claude to restart the broken line or table row and reopen any code block, without repeating earlier text\n\n**Unclear.** It is not clear whether this replaces an older continuation message, or whether a setting controls when it is used.\n\n**What**\n\nEach reply has a maximum length, called the output token limit. Tokens are the small chunks of text the model counts. Claude Code now contains a message for when a reply hits that limit and is cut off. It tells Claude that the rest will appear as a separate block, so Claude should:\n\n- restart the line, table row, list item or sentence that was cut off\n\n- reopen any code block or table header that was open\n\n- not repeat anything before that point\n\n- not comment on the interruption\n\n**Why**\n\nLong answers that hit the limit get joined back together. These instructions affect how code blocks and tables come out when that happens.\n\n- Area: System Reminders\n- Tier: You'll notice\n- Useful: 2\/5\n- Signal: 3\/5\n- Scope: individual\n- Heads-up: no"}