GPT-4o and GPT-4.1 API Retirement: The January 14 Migration Checklist

Description and use

By following the checklist, teams can migrate from the retiring GPT‑4o and GPT‑4.1 snapshots to gpt‑6‑luna, ensuring continued functionality, acceptable cost, and correct reasoning behavior before the January 14, 2027 shutdown.

Problem

OpenAI will retire GPT‑4o, GPT‑4o mini, GPT‑4.1 and GPT‑4.1 mini on January 14, 2027, requiring a migration to gpt‑6‑luna before the shutdown.

Outcome

By following the checklist, teams can migrate from the retiring GPT‑4o and GPT‑4.1 snapshots to gpt‑6‑luna, ensuring continued functionality, acceptable cost, and correct reasoning behavior before the January 14, 2027 shutdown.

Workflow

• 1. Verify any retiring snapshot is already dead (e.g., gpt‑4o‑2024‑05‑13) and note its replacement.
• 2. Grep all code, CI, Terraform, Docker, and environment files for the retiring model IDs.
• 3. Identify and update any fine‑tuned models, noting that gpt‑6‑luna does not support fine‑tuning.
• 4. For Azure users, adjust the migration timeline to the April 14, 2027 retirement date and plan for the earlier December 9, 2026 date for gpt‑4o‑2024‑05‑13.
• 5. Review the 92‑day notice and consider enterprise capacity requests if needed.
• 6. Understand gpt‑6‑luna’s reasoning defaults, tool‑calling API requirements, and parameter changes (temperature, top_p, logprobs).
• 7. Perform cost calculations for your traffic with reasoning off and on to assess savings.
• 8. Test gpt‑6‑luna on a copy of real traffic, evaluating output quality, format stability, and cost.
• 9. Update code to set reasoning_effort to none if a 4o‑style behavior is desired.
• 10. Deploy the updated system before mid‑December to allow a monitoring month in January.

Evidence

On January 14, 2027, five OpenAI snapshots stop answering: gpt‑4o‑2024‑08‑06, gpt‑4o‑2024‑11‑20, gpt‑4o‑mini‑2024‑07‑18, gpt‑4.1‑2025‑04‑14 and gpt‑4.1‑mini‑2025‑04‑14.

Source

Fine‑tuned versions of these models follow the same shutdown schedule as their underlying base models.

Source

OpenAI’s normal minimum is six months for generally available models. The October 14 notice gives 92 days.

Source

Luna is dramatically cheaper per token, but it does not behave like GPT‑4o or GPT‑4.1, and the differences are exactly the kind that pass a smoke test and fail in production.

Source