Skip to content

    © 2023 - 2025 OpenRouter, Inc

    Announcements

    Standardized finish reasons

    1/23/2025

    We've introduced finish_reason normalization to simplify the interpretation of responses:

    • Normalized Finish Reasons: Models now return standardized finish_reason values, aligned with OpenAI-style conventions.

    • Native Finish Reason: To retain model-specific details, the native_finish_reason is also returned.

    Why This Matters

    • Debugging and Understanding: Gain insight into a model's decision-making process.

    • Cross-Model Consistency: Streamline development with unified finish reasons across reasoning-capable models.