OpenAI has set a hard end date for GPT-4o and GPT-4.1 in its API. According to the company's API deprecations page, GPT-4o, GPT-4o mini, GPT-4.1, and GPT-4.1 mini will be removed from the API on January 14, 2027. OpenAI notified affected developers on October 14 and names a single replacement for all four: gpt-6-luna.
The notice has drawn fresh attention this week, and the timing matters for a second reason. A separate, larger batch of older models, announced back in April, shuts down this Friday, October 23.
What goes away on January 14
The October 14 entry lists five dated snapshots:
- GPT-4o:
gpt-4o-2024-08-06andgpt-4o-2024-11-20 - GPT-4o mini:
gpt-4o-mini-2024-07-18 - GPT-4.1:
gpt-4.1-2025-04-14 - GPT-4.1 mini:
gpt-4.1-mini-2025-04-14
Fine-tuned versions of these models "follow the same shutdown schedule as their underlying base models," the page says. Code that calls the plain gpt-4o alias is affected too: OpenAI's GPT-4o model page lists gpt-4o-2024-08-06 as the default snapshot behind that name.
That same model page still describes GPT-4o as OpenAI's "versatile, high-intelligence flagship model," a line that has not been updated to match the deprecation notice.
ChatGPT users lost GPT-4o earlier. OpenAI retired it from ChatGPT on February 13, 2026, saying at the time that there were "no changes" in the API. The January date closes that gap.
Three months, not six
The deprecations page also states OpenAI's own notice policy. Generally available models get "at least 6 months" of notice, specialized variants at least 3 months, and preview models can go in as little as two weeks. The policy allows shorter timelines when "safety or compliance concerns" require it.
The GPT-4 series notice runs three months, from October 14 to January 14. The entry does not say why it is shorter than the six-month minimum listed for generally available models, and it does not cite a safety or compliance reason. The page does say some developers "may be able to provision dedicated capacity for continued access after a model's shutdown date" by contacting sales.
The replacement is cheaper, but it is a different kind of model
GPT-6 Luna is listed at $0.10 per million input tokens and $0.50 per million output tokens on the Luna model page. GPT-4o is listed at $2.50 input and $10.00 output. At list price, Luna input costs one twenty-fifth as much and output one twentieth.
Two differences matter more than price for most migrations:
- Luna is a reasoning model. Its
reasoning.effortsetting accepts none, low, medium, high, xhigh, and max, with medium as the default. Teams moving latency-sensitive GPT-4o workloads will want to test the lower settings, since the default will spend tokens on reasoning that GPT-4o never did. - Luna does not support fine-tuning. The model page marks fine-tuning as "Not supported." That leaves teams with custom GPT-4o or GPT-4.1 models without a like-for-like target on OpenAI's list.
The fine-tuning problem is compounded by a separate schedule. OpenAI is winding down its self-serve fine-tuning platform, and on January 6, 2027, active existing customers "will no longer be able to create new fine-tuning jobs." That is eight days before the GPT-4 series shutdown. Inference on existing fine-tunes continues only until the base model is removed.
Friday's shutdown is larger
Before January arrives, the April 22 batch ends on October 23. It includes gpt-4-turbo, gpt-4-0613, gpt-4-1106-preview, gpt-3.5-turbo-0125, gpt-4.1-nano, the original gpt-4o-2024-05-13 snapshot, gpt-image-1, o1, o1-pro, o3-mini, and o4-mini. Fine-tuned GPT-3.5 Turbo, GPT-4, GPT-4.1 nano, o4-mini, babbage-002, and davinci-002 models go on the same day. OpenAI points most of these to the GPT-5.6 family: gpt-5.6-sol, gpt-5.6-terra, or gpt-5.6-luna.
The rest of the calendar is crowded as well:
- December 11, 2026: the August 2025 GPT-5, GPT-5 mini, and GPT-5 nano snapshots, plus
o3ando3-pro. - January 6, 2027:
tts-1,tts-1-hd, and twogpt-4o-mini-ttssnapshots, alongside the fine-tuning cutoff. - January 20, 2027: legacy audio and realtime families, including
gpt-4o-audioandgpt-4o-realtime. - February 26, 2027:
whisper-1and the GPT-4o transcription models. - April 1, 2027:
gpt-5.1,gpt-5.3-codex, andgpt-5.4-nano.
Why it matters
GPT-4o and GPT-4.1 are common defaults in older integrations, where model names tend to be hardcoded in config files, no-code workflows, and shared libraries. The January shutdown forces those apps onto a reasoning model with different latency, token use, and tooling defaults, and it removes the fine-tuning path that some of them depend on.
For teams that want to track this programmatically, OpenAI says the List models endpoint (GET /v1/models) returns an optional shutdown_date field for each model, so a scheduled check can flag any model in use that has a date attached.
