🔍 Read the full analysis: What Developers Need To Know About OpenAI Retiring GPT-4o, GPT-4.1, And GPT-5-nano on ThorstenMeyerAI.com
Get business pricing on tech for your team
- Business-only prices and quantity discounts
- Tax-exempt purchasing
- Multiple users, one account, clear invoices
TL;DR
OpenAI’s published deprecation schedule lists GPT-5 models and o3 for API retirement on Dec. 11, 2026, followed by GPT-4o and GPT-4.1 models on Jan. 14, 2027. The source says their listed replacements differ in price and use reasoning, which can affect billing and outputs. Some older models are reported to have stopped working on Oct. 23, 2026.
OpenAI plans to retire GPT-5 models and o3 from its API on Dec. 11, 2026, followed by GPT-4o and GPT-4.1 model snapshots on Jan. 14, 2027, according to the deprecation schedule described by ThorstenMeyerAI.com. The changes affect developers whose applications call those model IDs: they will need to move to a listed replacement and check for changes to price, model behavior and API settings before the shutdowns.
The schedule, which ThorstenMeyerAI.com says it checked on Nov. 1, sets Dec. 11 for gpt-5, gpt-5-mini, gpt-5-nano, gpt-5-pro, o3 and o3-pro. The article says 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 directed to gpt-5.6-sol with its pro reasoning mode.
On Jan. 14, 2027, the listed GPT-4o, GPT-4o mini, GPT-4.1 and GPT-4.1 mini snapshots are scheduled to leave the API. The source identifies gpt-6-luna as the recommended replacement for these models. It also reports that a number of older model IDs, including gpt-4.1-nano, gpt-4-turbo and the original GPT-4o snapshot, were shut down on Oct. 23. Developers still using those IDs may already be seeing failed requests, the article warns.
The article reports substantial price differences among the proposed replacements. At the short-context rates it cites, GPT-4o costs $2.50 per million input tokens and $10 per million output tokens, compared with $0.10 and $0.50 for gpt-6-luna. By contrast, the cited GPT-5 nano replacement costs $0.20 per million input tokens versus $0.05 for GPT-5 nano, while GPT-5 mini’s replacement is listed at $2.00 per million input tokens versus $0.25. The source says replacement models use reasoning, so hidden reasoning tokens are billed as output; developers may also need to remove some sampling parameters and review output limits.
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 Can Change Cost and Output
These retirements are not simply a matter of replacing one string in a configuration file. A migration can change token costs, response style and API settings, with consequences for both operating budgets and user-facing behavior. The source says the new models are reasoning models and that hidden reasoning tokens count as billed output. That means a cost estimate based only on visible response length may not match the eventual bill.
Pricing moves in different directions across the models cited. Some GPT-4-family users may see lower listed rates with gpt-6-luna, while the stated rates for GPT-5 nano and GPT-5 mini replacements are higher. These are the source’s quoted short-context prices, not a guarantee of an individual customer’s bill; actual usage, caching, context length and current pricing should be checked directly. Testing should cover representative prompts and output constraints, not just whether a request succeeds.
The source also flags a separate behavior change: OpenAI revised its public Model Spec on Oct. 29, removing much of its detailed style guidance, including a default preference for Markdown formatting. For developers who depend on a particular structure, the practical implication is to state that format explicitly in prompts and verify it in tests. The retirement schedule and the Spec revision are distinct developments, but together they make it prudent to recheck both model availability and application assumptions.
Products Worth Considering
Amazon Kindle Scribe Colorsoft, 64GB, 11-Inch Color Display with Premium Pen
Amazon Kindle Scribe (64GB) with Premium Pen, Metallic Jade
Jackery Explorer 2000 v2 Portable Power Station with Two 200W Solar Panels
As an affiliate, we earn on qualifying purchases.
Dates, Snapshots and Model Spec
The dates refer to API availability, not necessarily to all OpenAI products. ThorstenMeyerAI.com says OpenAI removed GPT-4o, GPT-4.1 and GPT-4.1 mini from ChatGPT on Feb. 13, 2026; the current schedule discussed in the article concerns apps and services using the API. Developers should not assume that a change to ChatGPT and an API shutdown have the same timing or scope.
The source says the deprecation page lists dated snapshots, while familiar short names such as gpt-4o and gpt-4.1 are also marked deprecated on their model pages. It advises treating those short names as affected by the January date. It reports that OpenAI says fine-tuned models follow the shutdown schedule of their base model and that the listed replacement models do not support fine-tuning.
The same article identifies Nov. 30, 2026, as the planned shutdown of OpenAI’s Evals platform, Agent Builder and reusable prompt objects. Existing evals reportedly became read-only on Oct. 31. These tools are separate from the model retirements, but the timing could affect teams relying on them to validate a migration. The source also describes an Oct. 29 Model Spec revision that emphasizes interpreting instructions by intent and removes much of the former style guidance.
“The replacements think before they answer.”
— ThorstenMeyerAI.com, summarizing the API migration
Open Questions on Notice and Access
The source says OpenAI’s deprecation page promises at least six months’ notice for generally available models unless safety or compliance concerns require a faster schedule. It also says the GPT-4-family notice was dated Oct. 14, giving about three months until the Jan. 14 retirement. The article does not establish which exception, if any, applies, so the reason for the shorter interval remains unclear.
It is also not clear from the source whether every affected customer has received notice, how any dedicated-capacity option would work, or whether a particular account could retain access beyond the listed dates. The source says OpenAI may offer some customers dedicated capacity after shutdown, but gives no eligibility criteria or terms. Replacement prices and model behavior can also change, so developers should verify current documentation rather than treating the cited Nov. 1 figures as permanent.
Finally, the article does not provide independent test results comparing the retired models with their replacements. A recommended replacement is not proof of equivalent output for a given application. Teams will need to assess their own prompts, latency, tool use, structured responses and quality requirements.
Audit IDs and Test Replacements
Developers should inventory model IDs across application code, configuration, scheduled jobs, third-party connectors and no-code automations, then match each one to the published dates. The immediate priority is to investigate any identifiers the source says stopped working on Oct. 23. For models scheduled to retire later, teams can begin testing the listed replacements in a staging environment before the relevant shutdown.
Tests should compare representative inputs and expected outputs, check structured-format and Markdown requirements, and measure latency and billed input and output tokens. Teams should also review unsupported sampling parameters and output-token limits; the source warns that a tight limit on a reasoning model can result in an empty response that is still billed. If a business depends on a particular model or has a committed-spend arrangement, the article recommends asking an OpenAI account team about the schedule and any continued-access options.
The next dated milestones in the source are the Nov. 30 tool shutdown, the Dec. 11 GPT-5 and o3 retirements, and the Jan. 14 GPT-4o and GPT-4.1 retirements. OpenAI’s deprecation page and pricing documentation are the places developers should check for updates before each migration decision.
Key Questions
When are GPT-5, GPT-4o and GPT-4.1 scheduled to leave the API?
The source reports that GPT-5 models and o3 are scheduled for Dec. 11, 2026. GPT-4o, GPT-4o mini, GPT-4.1 and GPT-4.1 mini snapshots are scheduled for Jan. 14, 2027.
Which replacements does the source identify?
It says OpenAI lists gpt-5.6-sol, gpt-5.6-terra and gpt-5.6-luna for the affected GPT-5 and o3 models, depending on the model. For the listed GPT-4o and GPT-4.1 snapshots, it identifies gpt-6-luna as the recommended replacement. Check OpenAI’s current documentation before changing production systems.
Will a replacement necessarily cost less?
No. The prices cited in the article vary by model: its listed GPT-4o-to-gpt-6-luna comparison is lower per token, while the cited GPT-5 nano and GPT-5 mini replacements have higher input rates. The figures are short-context prices reported as of Nov. 1, 2026, and should be rechecked.
Do fine-tuned models also get retired?
According to the source, OpenAI says fine-tuned models follow the shutdown schedule of their base model. The article also says the listed replacement models do not support fine-tuning, so teams using fine-tuned versions should confirm their migration options with OpenAI.
What should developers test before migrating?
Test real application prompts and expected output formats, along with token usage, latency, parameter compatibility and output limits. The source warns that reasoning tokens are billed as output and that a tight output limit can produce an empty, billed response.
Source: ThorstenMeyerAI.com
