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

Description and use

By following the checklist, teams can migrate to gpt‑6‑luna before the January 14 2027 shutdown, avoid runtime failures, maintain fine‑tuned workloads, and control costs despite the shorter notice period.

Problem

OpenAI is retiring GPT‑4o, GPT‑4o mini, GPT‑4.1 and GPT‑4.1 mini API snapshots on January 14 2027, leaving developers with a 92‑day notice and a new replacement model, gpt‑6‑luna, that behaves differently and does not support fine‑tuning.

Outcome

By following the checklist, teams can migrate to gpt‑6‑luna before the January 14 2027 shutdown, avoid runtime failures, maintain fine‑tuned workloads, and control costs despite the shorter notice period.

Workflow

• 1. Verify any retiring snapshot is already dead (e.g., gpt‑4o‑2024‑05‑13).
• 2. Grep all code, CI, Terraform, Docker, env files, and third‑party configs for the retiring snapshot IDs.
• 3. Identify fine‑tuned models and note they will expire with the base model.
• 4. For Azure users, note the April 14 2027 retirement dates and plan accordingly.
• 5. Understand the 92‑day notice and consider enterprise capacity if needed.
• 6. Recognize gpt‑6‑luna is a reasoning model; set reasoning_effort to none for 4o‑style behavior.
• 7. Recalculate costs using your traffic and reasoning settings.
• 8. Test gpt‑6‑luna on a copy of real traffic, evaluating quality, format stability, and cost.
• 9. Update code to use the Responses API if tool calls are required with reasoning.
• 10. Deploy changes by 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

The recommended replacements for this wave are the GPT-5.6 family (gpt-5.6-terra, gpt-5.6-sol, gpt-5.6-luna), not Luna, so do not blend the two migration lists.

Source

Set reasoning_effort to none if you want the closest thing to 4o-style behavior and cost.

Source

Against GPT-4o that is 25x cheaper on input and 20x on output.

Source