🔍 Read the full analysis: OpenAI Is Retiring GPT-4o, GPT-4.1, And GPT-5-nano In Its API: What To Do on ThorstenMeyerAI.com
Get movie nights delivered free with Prime
- Fast, free delivery on millions of items
- Prime Video, Amazon Music and more included
- Member-only deals all year
TL;DR
OpenAI’s published schedule says GPT-5, GPT-5 mini and GPT-5-nano leave the API on December 11, 2026; GPT-4o, GPT-4o mini, GPT-4.1 and GPT-4.1 mini follow on January 14, 2027. Developers should inventory model IDs, test the listed replacements, check cost and output behavior, and confirm dates with OpenAI because some details remain unclear.
OpenAI plans to retire several models from its API, with GPT-5, GPT-5 mini and GPT-5-nano scheduled to stop working on December 11, 2026, and GPT-4o, GPT-4o mini, GPT-4.1 and GPT-4.1 mini scheduled for January 14, 2027. The change affects developers and businesses using those model IDs in applications, automations and other API-connected services; OpenAI’s listed replacements may differ in price and behavior.
According to OpenAI’s deprecations page, the December 11 shutdown covers GPT-5, GPT-5 mini, GPT-5-nano, GPT-5-pro, o3 and o3-pro. OpenAI lists GPT-5.6 Sol as the replacement for GPT-5 and o3, GPT-5.6 Terra for GPT-5 mini, and GPT-5.6 Luna for GPT-5-nano. GPT-5-pro and o3-pro are mapped to GPT-5.6 Sol with its pro reasoning mode.
The January 14, 2027 date applies to listed snapshots of GPT-4o, GPT-4o mini, GPT-4.1 and GPT-4.1 mini. OpenAI recommends GPT-6 Luna for these models. The source article says the plain model names also show as deprecated, so teams should check their exact IDs rather than assuming a short name is unaffected.
Several other model IDs were already scheduled to stop working on October 23, 2026, including GPT-4.1-nano, the original GPT-4o snapshot, GPT-4-turbo, GPT-3.5-turbo, o1 and o4-mini. OpenAI’s notice also says fine-tuned versions follow the shutdown schedule of their base model. The source reports that the listed replacements do not support fine-tuning.
Infographic 01 · The shutdown calendar
What stops working, and when
OpenAI's API deprecations, as listed on 1 November 2026. GPT-4o and GPT-4.1 already left ChatGPT on 13 February 2026; this round is about apps built on the API.
- 23 Oct 2026Already offAlready shut down
gpt-3.5-turbo,gpt-4,gpt-4-turbo,gpt-4.1-nano,gpt-4o-2024-05-13,o1,o1-pro,o3-mini,o4-mini,gpt-image-1. Calls to these fail today. - 30 Nov 2026Your testsEvals platform, Agent Builder, reusable prompts
Not models, but tools many teams test with. Existing evals became read-only on 31 Oct. Export them first.
- 11 Dec 2026Notified 11 JunGPT-5 family and o3
gpt-5→gpt-5.6-sol·gpt-5-mini→gpt-5.6-terra·gpt-5-nano→gpt-5.6-luna·o3→gpt-5.6-sol·gpt-5-pro,o3-pro→gpt-5.6-sol(pro mode) - 14 Jan 2027Notified 14 OctGPT-4o, GPT-4o mini, GPT-4.1, GPT-4.1 mini
Snapshots
gpt-4o-2024-08-06,gpt-4o-2024-11-20,gpt-4o-mini-2024-07-18,gpt-4.1-2025-04-14,gpt-4.1-mini-2025-04-14→gpt-6-luna. Fine-tuned versions follow their base model. - Jan to Apr 2027Plan aheadAudio, speech, older GPT-5.x
6 Jan:
tts-1and others · 20 Jan: older audio and realtime models · 26 Feb:whisper-1, GPT-4o transcription · 1 Apr:gpt-5.1,gpt-5.4-nano,gpt-5.3-codex
Sources: OpenAI deprecations page (checked 1 Nov 2026) · OpenAI Help Center, model release notes. Dates as listed by OpenAI; shutdowns can be rescheduled.
Powered by Thorsten Meyer AI
Migration Affects Cost and Output
A retired model ID can break API calls embedded in software, internal tools and no-code workflows. The effect is not limited to a developer changing one line: teams may need to find model references in configuration files, test fixtures, prompts, and services maintained by different groups. Any unupdated request to a shut-down ID may fail, so identifying dependencies before the listed dates is a practical reliability task.
Changing models can also alter budgets. The source article, citing OpenAI’s pricing page as checked November 1, reports that GPT-6 Luna is listed at $0.10 per million input tokens and $0.50 per million output tokens, compared with GPT-4o at $2.50 and $10.00. But the move from GPT-5-nano to GPT-5.6 Luna raises the listed input price from $0.05 to $0.20 per million tokens and output price from $0.40 to $1.20. GPT-5 mini’s listed replacement, GPT-5.6 Terra, is also more expensive per token than GPT-5 mini. These figures use short-context pricing and should be checked again before budgeting.
The replacements are described as reasoning models. The source says hidden reasoning tokens are billed as output, and some sampling parameters must be removed. It also warns that a low output-token limit could result in an empty response that is still billed. Compatibility testing needs to cover both technical settings and response quality, not just whether a request succeeds.
Popular right now
Breville Bambino Espresso Machine, Brushed Stainless Steel (BES450BSS)
Vitamix Propel Series 510 Professional-Grade Blender, 3 Programs, Black
KitchenAid Classic Series 4.5-Quart Tilt-Head Stand Mixer, Onyx Black
As an affiliate, we earn on qualifying purchases.
The API Schedule and Earlier Changes
This is an API retirement schedule, not a new announcement that those models have been removed from ChatGPT. The source reports that OpenAI removed GPT-4o, GPT-4.1 and GPT-4.1 mini from ChatGPT on February 13, 2026; the dates in this article concern access through the API, which developers use to build their own products and workflows.
OpenAI’s deprecations page is the basis for the dates reported here. The source says OpenAI notified customers about the December retirements on June 11 and the January retirements on October 14, and that the company emails affected customers. It also reports that OpenAI’s stated general policy is to provide at least six months’ notice for generally available models, unless safety or compliance concerns require a faster timeline. The January date follows a three-month interval from the notification date; the listed notice does not explain whether an exception applies.
There are nearby changes that may affect migration work. The source says OpenAI scheduled its Evals platform, Agent Builder and reusable prompt objects at v1/prompts to shut down on November 30, 2026, with existing evals read-only from October 31. Those services are distinct from the model retirements but could be part of a team’s testing or prompt-management process.
Notice Period and Model Behavior
Some details remain unresolved. The source does not explain why the January 14, 2027 shutdown was listed three months after notification despite the reported six-month notice policy. OpenAI’s deprecation entry may provide additional information, and customers with contracts or committed spend are advised in the source to ask their account teams about their arrangements.
The precise effect of switching models will depend on each application. The source reports that OpenAI revised its public Model Spec on October 29, removing much of its detailed style guidance, including a default preference for Markdown, and directing models to interpret instructions by intent. That does not establish how any particular API response will change. Developers need to test their own prompts and required formats; the source also says OpenAI acknowledges that models do not yet fully reflect the Spec.
The source reports that some customers may be able to purchase dedicated capacity for continued access after a shutdown, but it does not provide eligibility terms, prices or availability. Those details should be confirmed directly with OpenAI.
Audit IDs Before December
Start by searching code, configuration, deployment settings, prompt tools and automation platforms for every model ID in use. Record the exact ID, the service that calls it, the owner, and whether it is a dated snapshot or a short name. Prioritize models already past their October 23 shutdown date, then plan separate migrations for December 11 and January 14.
Next, route representative requests to the replacement model in a test environment. Compare correctness, formatting, latency, token use and total cost against the existing setup. Review sampling parameters and output-token limits for compatibility with reasoning models, and check any fine-tuning dependency against the announced replacement options. Include empty-output cases and the formats your application relies on in the test plan.
Recheck OpenAI’s current deprecation and pricing pages before each migration, since dates and prices can change. The source also reports a November 30 shutdown for Evals, Agent Builder and reusable prompt objects, so teams that use those services should account for their separate deadline. OpenAI has not provided a universal migration procedure in the source material; customers should confirm account-specific access or contract questions with their OpenAI contact.
Key Questions
When will GPT-4o stop working in the OpenAI API?
OpenAI’s schedule lists January 14, 2027 for the named GPT-4o snapshots, including GPT-4o-2024-08-06 and GPT-4o-2024-11-20. The source says the plain GPT-4o name is also marked deprecated, so developers should verify the exact ID they use.
When are GPT-5-nano and GPT-5 mini being retired?
The listed shutdown date for GPT-5-nano and GPT-5 mini is December 11, 2026. The replacements named in the schedule are GPT-5.6 Luna for GPT-5-nano and GPT-5.6 Terra for GPT-5 mini.
What should developers do before their model is retired?
Find every affected model ID, test the recommended replacement with real representative requests, and compare response quality, format, latency and cost. Check parameters and output limits as well, since the replacements are described as reasoning models.
Will the replacement models cost less?
Not in every case. The source’s November 1 pricing comparison shows lower listed token prices for GPT-6 Luna than GPT-4o, but higher listed prices for GPT-5.6 Luna than GPT-5-nano and for GPT-5.6 Terra than GPT-5 mini. Pricing depends on model and token type, so check OpenAI’s current pricing before estimating spend.
Can teams keep using fine-tuned versions?
The source says fine-tuned models follow the shutdown schedule of their base model and that the listed replacements do not support fine-tuning. Teams relying on fine-tuning should confirm available options with OpenAI rather than assume their existing model will remain accessible.
Source: ThorstenMeyerAI.com
