OpenRouter：Announcements· OpenRouter·· 2025-01-23AI 评分45
Standardized finish reasons
Standardized finish reasons
AI 导读
模型现在开始返回标准化的完成原因值，并与OpenAI风格的惯例保持一致。这一变化意味着不同模型（如GPT、Claude、LLaMA）的输出终止原因将采用统一的分类和表述方式，例如“stop”、“length”或“content_filter”，从而提升API调用的可预测性和跨平台兼容性。开发者无需再为不同模型适配各异的终止逻辑，简化了集成与调试流程。
正文 · 原文
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.
来源：OpenRouter：Announcements · openrouter.ai