What changed
On October 14, 2026, OpenAI notified API developers that GPT-4o snapshots gpt-4o-2024-08-06 and gpt-4o-2024-11-20, GPT-4o mini gpt-4o-mini-2024-07-18, GPT-4.1 gpt-4.1-2025-04-14 and GPT-4.1 mini gpt-4.1-mini-2025-04-14 will be removed from the API on January 14, 2027. Fine-tuned versions built on these bases follow the same shutdown date. OpenAI recommends gpt-6-luna for each. This is an API retirement, not the February 2026 ChatGPT model-picker retirement. A separate older snapshot set, including gpt-4o-2024-05-13 and gpt-4.1-nano-2025-04-14, has an earlier October 23, 2026 cutoff under an April notice.
Why it matters
Applications that pin old model IDs can stop serving requests at the cutoff even if the rest of their stack is healthy. The three-month notice leaves limited time to inventory SDK defaults, scheduled jobs, fine-tuned endpoints and third-party integrations. A replacement model name alone is not a regression test: tool calling, structured outputs, reasoning settings, latency, long-context behavior and usage bills need checking against production workloads. Small SaaS teams and integrators are especially exposed when model IDs are embedded in old code or customer configurations.
Which model snapshots stop on January 14
OpenAI's deprecations table lists five dated snapshots: 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. Fine-tuned descendants of those models have the same January 14, 2027 retirement. OpenAI names gpt-6-luna as the recommended replacement for each. Check the exact model ID returned or requested by your integration rather than relying on a family nickname.
A separate October 23 deadline comes first
The April 22 legacy-model notice already scheduled gpt-4o-2024-05-13, gpt-4.1-nano-2025-04-14 and other legacy models for October 23, 2026. The January notice does not extend that earlier deadline. The earlier February 13, 2026 removal of GPT-4o and GPT-4.1 from ChatGPT did not itself remove API access; these are separate products and schedules.
What to test before changing production traffic
Inventory explicit model IDs in code, environment variables, database configuration, batch jobs and SDK wrappers. Include fine-tuned model references and vendor-managed integrations. Compare the proposed gpt-6-luna replacement on representative production inputs: tool invocation and function-call parameters, JSON/schema validation, refusals, long-context outputs, response latency and token use. Confirm API endpoint support and reasoning settings from the current model documentation before switching; Chat Completions tool-call behavior may differ when reasoning is enabled. Roll out gradually with error, quality and cost monitoring.
Evidence boundaries
The official API deprecations page establishes the dates and replacement recommendation. Independent developer reporting repeats the cutoff and highlights migration compatibility risks, but does not prove that every application will break or quantify the number of affected developers. OpenAI has not announced an extension. Current list pricing and endpoint features can change before the deadline, so teams should recheck official model and pricing documentation during migration.