Source: Messages API reference (Claude API)

Reference for Claude's Create Message endpoint, including max_tokens, stop_sequences, stop_reason values and usage accounting.

Contents
  • What: official API reference
  • Author: Anthropic
  • Date: accessed 2026-11-14 (API version 2026-10-21)
  • Original: Messages API

Summary

The reference lists request parameters and response fields for POST /v1/messages. Relevant here are the length and stopping controls and how usage is reported.

Key claims

  • max_tokens is the maximum to generate; the model may stop earlier; 0 populates the prompt cache without generating.
  • stop_sequences are custom strings that end generation; the matched string is returned in stop_sequence.
  • stop_reason values: end_turn, max_tokens, stop_sequence, tool_use, pause_turn, refusal, model_context_window_exceeded.
  • Usage token counts do not map one-to-one to visible content (requests are transformed and outputs parsed); output_tokens is non-zero even for an empty reply.
  • Total input = input_tokens + cache_creation_input_tokens + cache_read_input_tokens.
  • The 2026-10-21 reference has no temperature, top_p or top_k parameters.

Feeds

Sampling parameters · Context window

Suggest changes →