pricingopenaideprecations

OpenAI Shuts Down GPT-4o and GPT-4.1 on January 14, 2027

Short answer. OpenAI shuts down GPT-4o, GPT-4o mini, GPT-4.1 and GPT-4.1 mini on January 14, 2027, according to its deprecations page, with a notice date of October 14, 2026. It names GPT-6 Luna as the replacement for all four. Luna costs $0.10 input and $0.50 output per million tokens; GPT-4o costs $2.50 and $10.00.

In plain words. Four older OpenAI models stop working in mid-January. OpenAI points everyone to one newer model, GPT-6 Luna. Luna is cheaper than all four. A token is a short piece of text, about three-quarters of a word. Prices are per million tokens.

Read on November 16, 2026. We will check this page again before January 14, 2027.

The four rows

OpenAI’s deprecations page lists upcoming shutdowns in a table with the columns “Shutdown date | Affected model | Status | Recommended replacement | Notice date”. On November 16, 2026 it printed these four rows:

  • “Jan 14, 2027 | GPT-4o ( gpt-4o-2024-08-06 , gpt-4o-2024-11-20 ) | Deprecated | gpt-6-luna | Oct 14, 2026 |”
  • “Jan 14, 2027 | GPT-4o mini ( gpt-4o-mini-2024-07-18 ) | Deprecated | gpt-6-luna | Oct 14, 2026 |”
  • “Jan 14, 2027 | GPT-4.1 ( gpt-4.1-2025-04-14 ) | Deprecated | gpt-6-luna | Oct 14, 2026 |”
  • “Jan 14, 2027 | GPT-4.1 mini ( gpt-4.1-mini-2025-04-14 ) | Deprecated | gpt-6-luna | Oct 14, 2026 |”

So the rows cover five snapshots: two for GPT-4o and one for each of the other three. The same page defines the two states that matter here: “Deprecated. OpenAI has announced the model’s retirement. You can still use it until its shutdown date.” and “Shutdown. The model no longer accepts requests through the standard API.”

What the page does not say matters as much. The rows name snapshots, not the short names. GPT-4o’s model page lists “all available snapshots and aliases for GPT-4o”, with gpt-4o among them, but neither page says what happens to the alias gpt-4o on January 14. Neither says that requests will be routed to GPT-6 Luna either. Google’s deprecations page does say that for Gemini 2.5 Pro; OpenAI’s does not. Until OpenAI says otherwise, the safe reading is the page’s own definition: after the shutdown date, requests are not accepted.

The page also states its notice policy: “Generally available models: At least 6 months.” From October 14, 2026 to January 14, 2027 is three months. The policy opens with an exception, “Unless safety or compliance concerns require a faster timeline”, and the page does not say whether it applies here; we do not guess. It adds that “developers may be able to provision dedicated capacity for continued access after a model’s shutdown date”, through OpenAI’s sales team.

This is about the API. The pages we read say nothing about the ChatGPT app, and neither do we.

What the four cost against GPT-6 Luna

The prices come from the Markdown version of OpenAI’s pricing page, read on November 16, 2026. The four old models sit in a table with the columns “Model | Input | Output | Cached input |”: “gpt-4o | $2.50 | $10.00 | $1.25 |”, “gpt-4o-mini | $0.15 | $0.60 | $0.075 |”, “gpt-4.1 | $2.00 | $8.00 | $0.50 |” and “gpt-4.1-mini | $0.40 | $1.60 | $0.10 |”. GPT-6 Luna’s row, with short-context prices first and long-context prices after, is “gpt-6-luna | $0.10 | $0.50 | $0.01 | $0.125 | $0.25 | $0.20 | $0.75 | $0.02 | $0.25 | $0.50 |”.

Take a month of 1,000,000 requests, each with 2,000 input tokens and 500 output tokens, without caching: 2 billion input tokens and 500 million output tokens.

Model, per million tokensInputOutputCached input1M requests, 2,000 in + 500 out
GPT-4o$2.50$10.00$1.25$10,000.00
GPT-4.1$2.00$8.00$0.50$8,000.00
GPT-4.1 mini$0.40$1.60$0.10$1,600.00
GPT-4o mini$0.15$0.60$0.075$600.00
GPT-6 Luna$0.10$0.50$0.01$450.00

Every one of the four costs more than Luna. GPT-4o’s month costs 22× Luna’s, GPT-4.1’s 18×, GPT-4.1 mini’s 3.6× and GPT-4o mini’s 1.3×. The gap is widest on cached input: GPT-4o’s $1.25 is 125× Luna’s $0.01, and GPT-4o mini’s $0.075 is 7.5× it. GPT-4o mini is the closest of the four, 1.5× Luna on input and 1.2× on output.

A long prompt: where Luna changes price

Luna has a second price band. Its model page says: “Prompts with more than 272K input tokens are priced at 2x input and cache rates and 1.5x output for the full request.” Above that line Luna costs $0.20 input, $0.75 output and $0.02 cached input, as its row above prints.

GPT-4o and GPT-4o mini cannot take such a prompt: their model pages give “128,000 context window”. GPT-4.1 and GPT-4.1 mini can, with “1,047,576 context window”, and they have one price at any length. So take 10,000 requests a month of 300,000 input tokens and 2,000 output tokens, without caching: 3 billion input tokens and 20 million output tokens.

ModelInputOutput10,000 requests, 300,000 in + 2,000 out
GPT-4.1$2.00$8.00$6,160.00
GPT-4.1 mini$0.40$1.60$1,232.00
GPT-6 Luna, prompts over 272K$0.20$0.75$615.00

Here GPT-4.1 costs 10× Luna and GPT-4.1 mini 2.0×. The band narrows the gap but does not close it: Luna’s long-prompt input, $0.20, is still half of GPT-4.1 mini’s $0.40.

Context and output length

From the model pages, read on November 16, 2026:

ModelContext windowMax output tokens
GPT-4o128,00016,384
GPT-4o mini128,00016,384
GPT-4.11,047,57632,768
GPT-4.1 mini1,047,57632,768
GPT-6 Luna1,050,000128,000

Luna’s page gives “1,050,000 context window” and “128,000 max output tokens”, so it takes longer prompts than GPT-4o and slightly longer than GPT-4.1, and its longest answer is 3.9× GPT-4.1’s. Its page describes it as “our most efficient model for focused, high-volume tasks.” How its answers compare with GPT-4o’s is something only a test on real requests can show; OpenAI’s own migration advice says “Run representative requests against both models.”

OpenAI’s other shutdowns on record

Our deprecations page holds every dated OpenAI shutdown. Around this one:

Shutdown dateModelsRecommended replacement
October 23, 2026o1, o1-pro, o3-mini, GPT-4, GPT-4 Turbo, GPT-3.5 Turbogpt-5.6-sol; gpt-5.6-terra for GPT-3.5 Turbo
December 11, 2026GPT-5, GPT-5 mini, GPT-5 nano, GPT-5 Pro, o3, o3-progpt-6.1-sol; gpt-6.1-terra for GPT-5 mini; gpt-6-luna for GPT-5 nano
January 14, 2027GPT-4o, GPT-4o mini, GPT-4.1, GPT-4.1 minigpt-6-luna
February 1, 2027GPT-5.5 Cyber (gpt-5.5-cyber-preview)none named
April 1, 2027GPT-5.3 Codex, GPT-5.1, GPT-5.4 nanogpt-6-sol; gpt-6-luna for GPT-5.4 nano

The October 23 shutdown is covered in its own post, and the changes of December 1 in What Changes on December 1, 2026.

Our correction

Until November 16, 2026, AI Signal listed these four models as deprecated with no shutdown date. That was wrong. OpenAI’s page had printed “Jan 14, 2027” for them since its notice of October 14, 2026, and it was already there in our reading of October 26. OpenAI had changed the table’s format, with a status column and model names like “GPT-4.1 ( gpt-4.1-2025-04-14 )”, and our reader did not pick up those rows. We fixed the reader on November 16, added the date to the four rows, and logged the error as our own correction, not a change by OpenAI, in the price-changes log.

How we checked

On November 16, 2026 we read OpenAI’s deprecations page and the Markdown version of its pricing page by script, and the model pages of GPT-4o, GPT-4o mini, GPT-4.1, GPT-4.1 mini and GPT-6 Luna. Every quote above is from those readings. The four rows are in our dataset with the shutdown date and GPT-6 Luna as the replacement, and the weekly check reads the deprecations page every Monday. The monthly bills are our own arithmetic from the per-million prices, in US dollars, on the Standard tier, without caching or Batch; ratios are the larger bill divided by the smaller.

Frequently asked questions

When does OpenAI shut down GPT-4o and GPT-4.1?

On January 14, 2027: OpenAI's deprecations page, read on November 16, 2026, prints 'Jan 14, 2027 | GPT-4o ( gpt-4o-2024-08-06 , gpt-4o-2024-11-20 ) | Deprecated | gpt-6-luna | Oct 14, 2026 |', and the same date for GPT-4o mini, GPT-4.1 and GPT-4.1 mini, with October 14, 2026 as the notice date.

What should I migrate GPT-4o to?

OpenAI's deprecations page names one recommended replacement for all four models: 'gpt-6-luna'. On OpenAI's pricing page, read on November 16, 2026, GPT-6 Luna costs $0.10 per million input tokens and $0.50 per million output tokens, against $2.50 and $10.00 for GPT-4o.

Will requests to gpt-4o be routed to GPT-6 Luna after January 14, 2027?

OpenAI's deprecations page does not say so: it defines a shut-down model as one that 'no longer accepts requests through the standard API.' The four rows name dated snapshots, such as gpt-4o-2024-08-06, and the page does not say what happens to the alias gpt-4o.

Drafted with AI assistance from the cited source and reviewed by Álvaro Lucero Rufino, who is responsible for what stays published here. Automated publishing was discontinued in September 2026; posts that no longer met that standard were removed rather than left up. This is original explanatory writing, not a reproduction — read the source above for the full story.