AI Info Guide
AI news for everyone

OpenAI sets January 14 API shutdown for specified GPT-4o and GPT-4.1 models

October 15, 2026 · Central Time

OpenAI's October 14, 2026 deprecation notice schedules removal of specified GPT-4-series models from its API on January 14, 2027. The notice names particular model snapshots rather than announcing that all OpenAI models will stop working. Its shutdown schedule is an API lifecycle notice, not a statement about the availability of every ChatGPT product.

The affected GPT-4o snapshots are gpt-4o-2024-08-06 and gpt-4o-2024-11-20. The table also lists gpt-4o-mini-2024-07-18. Recording the full deployed identifier matters because a family name alone does not tell a team whether its current application uses one of the snapshots in the notice.

The same shutdown date applies to gpt-4.1-2025-04-14 and gpt-4.1-mini-2025-04-14. That makes five named snapshots across four table rows. The notice does not list every model with a GPT-4-related name, so teams should compare their actual identifiers with the documented entries instead of treating the label as a universal cutoff.

Fine-tuned versions of the listed models follow the same shutdown schedule as their underlying base models, according to OpenAI. A customized model therefore is not exempt merely because it has a separate fine-tuned identifier. An inventory should include those dependent versions as well as direct calls to the base snapshots.

OpenAI recommends gpt-6-luna as the replacement in each row. A recommended replacement is not evidence that an application's prompts, tools or output requirements will behave identically after a name change. The deprecation table does not supply a cost comparison, latency result or compatibility guarantee for a particular workload.

Developers should use the exact identifiers and shutdown date to plan their own migration checks. The source supports a future January 14 removal, not a claim that these APIs have already shut down in October. Price assumptions, endpoint compatibility and any retraining plan need separate current documentation and testing before a production switch.

Original sources

↑ Back to top