OpenAI发布了关于如何从GPT-4.x模型迁移至GPT-6 Luna模型的指导文档。
该文档建议开发者将当前基于GPT-4.x的应用场景逐步转移至GPT-6 Luna平台。
OpenAI Publishes Migration Framework from GPT-4.x to GPT-6 Luna OpenAI has published migration guidance advising developers to transition existing GPT-4.x use cases to GPT-6 Luna.
资料显示，相较于需迁移的GPT-4.x系列模型，GPT-6 Luna具备更强的功能表现且单位令牌成本更低。
According to the documentation, GPT-6 Luna offers higher capabilities at a lower per-token price compared to the GPT-4.x models covered by the transition.
OpenAI建议开发者首先将推理强度参数设为无状态模式进行测试，随后再依据延迟表现、输出质量及成本因素进行相应调整。
OpenAI recommends that developers initially test the model with reasoning effort set to none before adjusting for latency, quality, and cost.
整个迁移流程分为四个步骤：确立性能基准值、针对各类特殊场景及工具调用情况评估Luna模型表现、优化提示词设置与输出详略程度，最后分阶段将新模型投入实际生产环境。目前针对普通用户及企业级客户的技术支持渠道均已正式开通。
The guide outlines a four-step implementation process: establishing a performance baseline, evaluating Luna across edge cases and tool use, tuning prompts and verbosity, and executing a gradual production rollout. Support channels have been opened for both self-serve and enterprise customers.