OpenAI Is Switching Off API Models: How To Find And Fix What Breaks

🔍 Read the full analysis: OpenAI Is Switching Off API Models: How To Find And Fix What Breaks on ThorstenMeyerAI.com

Before you orderOffer from Amazon

Get the latest gadgets delivered free with Prime

• Fast, free delivery on millions of items
• Prime Video, Amazon Music and more included
• Member-only deals all year

Start your free Prime trial →Free trial for eligible customers · Cancel anytime

As an affiliate, we earn on qualifying purchases.Hide

OpenAI’s API deprecation schedule lists model shutdowns already dated Oct. 23, 2026, and future retirements on Dec. 11 and Jan. 14, 2027. The changes affect developers and businesses running apps on those model IDs, who may need to update configurations and test replacement models before calls stop working or outputs change.

According to the deprecation page as described by ThorstenMeyerAI.com, models listed as shut down on Oct. 23 include GPT-3.5 Turbo, GPT-4, GPT-4 Turbo, GPT-4.1 Nano, the original GPT-4o snapshot, o1, o1-pro, o3-mini and o4-mini. The source also lists the image model gpt-image-1. Any applications still calling those IDs may fail, though the source does not report how many users or services are affected.

The next model retirements are scheduled for Dec. 11, 2026: GPT-5, GPT-5 mini, GPT-5 nano, GPT-5 pro, o3 and o3-pro. OpenAI’s listed replacements include models in the GPT-5.6 family. On Jan. 14, 2027, the schedule lists GPT-4o, GPT-4o mini, GPT-4.1 and GPT-4.1 mini snapshots, with gpt-6-luna named as the replacement. OpenAI says fine-tuned versions follow their base model’s shutdown schedule; the source says the replacements do not support fine-tuning.

The same source reports that OpenAI plans to shut down its Evals platform, Agent Builder and reusable prompt objects on Nov. 30, 2026. Existing evals became read-only on Oct. 31, it says. The source also describes a revised Model Spec dated Oct. 29, which removes much of the detailed style guidance, including a default to format answers in Markdown.

Migration Risks for API Customers

Model retirement is not just a matter of changing a string in a configuration file. A replacement can affect cost, latency, output format and application behavior, so teams need to test the specific workflows that rely on a retiring model. The source says the replacements are reasoning models and that hidden reasoning tokens are billed as output; it also warns that certain sampling parameters may need to be removed and that a low output-token limit could result in an empty, billed response.

Prices can move in either direction. Using short-context prices per million tokens, the source lists GPT-4o at $2.50 input and $10 output, compared with $0.10 and $0.50 for gpt-6-luna. For GPT-5 nano’s listed replacement, GPT-5.6 Luna, it gives $0.05 input and $0.40 output versus $0.20 and $1.20. GPT-5 mini’s move to GPT-5.6 Terra is listed as $0.25 and $2 versus $2 and $12. These figures are the source’s snapshot checked Nov. 1, 2026, not a guarantee of future prices.

The Model Spec change adds another reason to make requirements explicit. If a product needs JSON, Markdown, a particular tone or a fixed response structure, developers may need to specify and validate that behavior rather than rely on a general style default. The source quotes OpenAI as saying its models “do not yet fully reflect” the Spec, so the document should not be treated as a guarantee of current model behavior.

Products Worth Considering

Jackery Solar Generator 1000 v2 with 200W Solar Panel

AmazonView Latest Price

LG UltraGear 32GX850A-B 32-inch 4K UHD Glossy OLED Gaming Monitor

AmazonView Latest Price

Apple Watch SE 3 GPS 40mm Smartwatch

AmazonView Latest Price

As an affiliate, we earn on qualifying purchases.

The Schedule and Earlier Changes

The source says OpenAI announced the Oct. 23 retirements in April, the Dec. 11 set on June 11, and the Jan. 14 models on Oct. 14. It also notes that ChatGPT users lost access to GPT-4o, GPT-4.1 and GPT-4.1 mini in February 2026. Those ChatGPT changes are separate from the current schedule, which concerns the API and services built on it.

OpenAI’s deprecation page, as summarized by ThorstenMeyerAI.com, says it emails affected customers and documents model retirements there. The source further reports that OpenAI promises at least six months’ notice for generally available models unless safety or compliance concerns require a faster timeline. It points out that the stated interval from Oct. 14 to Jan. 14 is three months; the notice entry does not explain which exception, if any, applies.

The source also lists later 2027 dates: some text-to-speech models on Jan. 6, older audio and realtime models on Jan. 20, Whisper and GPT-4o transcription models on Feb. 26, and GPT-5.1, GPT-5.4 Nano and GPT-5.3 Codex on April 1. These dates and model names are reported from the schedule checked Nov. 1, 2026.

Open Questions on Timing and Behavior

The schedule gives dates and suggested replacements, but the source does not establish whether every customer has received an email or whether all listed replacements will behave equivalently in every application. It also does not explain why the Jan. 14 retirement notice falls three months after the Oct. 14 notice despite the reported six-month notice policy; the applicable exception is unclear.

Pricing is subject to change, and the cited comparisons use short-context rates for prompts up to 272,000 tokens. The source does not provide a complete cost estimate for specific workloads, nor does it quantify how reasoning-token billing or changed outputs would affect each customer’s bill. Developers should verify current documentation and prices against their own usage.

Audit Models Before Each Deadline

Teams using the API should search configuration files, code, test fixtures, automation platforms and other deployment settings for both dated snapshots and short model names. The source warns that aliases such as gpt-4o and gpt-4.1 should be checked alongside the dated IDs, since the model pages mark them deprecated.

Before switching, run representative requests against the proposed replacement and compare output quality, formatting, latency, token use and total cost. Check whether the application relies on fine-tuning or parameters the replacement does not support, and give output limits particular attention. OpenAI’s published schedule and pricing should be checked again as deadlines approach; customers with committed spend or questions about notice periods can ask their OpenAI account team.

The next listed milestones are the Nov. 30 shutdown of Evals, Agent Builder and reusable prompts, followed by the Dec. 11 model retirements. The source does not report a separate migration deadline or testing tool beyond those dates.

Key Questions

Which OpenAI API models are scheduled to shut down on Dec. 11, 2026?

The source lists GPT-5, GPT-5 mini, GPT-5 nano, GPT-5 pro, o3 and o3-pro. It says the listed replacements are in the GPT-5.6 family.

When are GPT-4o and GPT-4.1 API snapshots scheduled to retire?

The source lists relevant GPT-4o, GPT-4o mini, GPT-4.1 and GPT-4.1 mini snapshots for Jan. 14, 2027, with gpt-6-luna named as the replacement.

Do fine-tuned versions keep working after their base model is retired?

According to the source’s summary of OpenAI’s notice, fine-tuned models follow the shutdown schedule of their base model. The source says the listed replacement models do not support fine-tuning.

Why should developers test a replacement instead of only changing the model ID?

The source describes differences in pricing, reasoning-token billing, supported parameters and output behavior. Testing helps identify application-specific changes, including formatting or responses affected by output limits.

Source: ThorstenMeyerAI.com