OpenAI shuts down GPT-3.5 Turbo, GPT-4, o1, o3-mini and o4-mini in its API on October 23

What shuts down on October 23

OpenAI announced on April 22, 2026 that it would retire a batch of older API models, and October 23 is the date on its deprecations page. The page says what that means: “At the time of the shut down, the model or endpoint will no longer be accessible.” It gives a day, not a time.

Shuts down, with its other names	OpenAI’s substitute
gpt-3.5-turbo, gpt-3.5-turbo-0125	gpt-5.6-terra
gpt-4, gpt-4-0613	gpt-5.6-sol
gpt-4-turbo, gpt-4-turbo-2024-04-09, gpt-4-1106-preview	gpt-5.6-sol
gpt-4o-2024-05-13	gpt-5.6-sol
gpt-4.1-nano, gpt-4.1-nano-2025-04-14	gpt-5.6-luna
o1, o1-2024-12-17	gpt-5.6-sol
o1-pro, o1-pro-2025-03-19	gpt-5.6-sol in pro mode
o3-mini, o3-mini-2025-01-31	gpt-5.6-sol
o4-mini, o4-mini-2025-04-16	gpt-5.6-terra
gpt-image-1	gpt-image-2.5-sunburst or gpt-image-2.5-flare

The -completions names of GPT-3.5 Turbo, GPT-4 and GPT-4 Turbo go too. So do fine-tuned models built on GPT-3.5 Turbo, GPT-4, GPT-4.1 nano and o4-mini, and on the older babbage-002 and davinci-002.

What keeps working, for now

The names are close enough to catch people out.

• gpt-4o keeps working. Only the May 2024 snapshot, gpt-4o-2024-05-13, goes on October 23. The gpt-4o alias points to gpt-4o-2024-08-06, according to its model page.
• GPT-4o, GPT-4o mini, GPT-4.1 and GPT-4.1 mini have until January 14, 2027. OpenAI notified their users on October 14 and names gpt-6-luna as the replacement for all four.
• o3 and the first GPT-5 models, o3-2025-04-16, o3-pro-2025-06-10 and the first GPT-5 snapshots, go on December 11, 2026.
• This is the API. The deprecations page “lists all API deprecations,” and nothing in this notice mentions ChatGPT or its plans.
• Azure OpenAI follows Microsoft’s schedule. Microsoft’s retirement schedule retires o1, o1-pro, o3-mini and o4-mini on November 19, 2026, and gpt-4o version 2024-05-13 on December 9, 2026, when Microsoft says it upgrades those deployments to gpt-5.6-sol automatically on the Standard SKU. Azure’s gpt-4.1-nano lasts until April 14, 2027.

Who it catches

It’s rarely the code you wrote last month. It’s the default nobody changed.

• Libraries and plugins with a default model. InternLM’s lagent agent framework has an issue about defaults set to gpt-3.5-turbo and gpt-4o-2024-05-13. Pressidium’s cookie-consent plugin closed the same problem on October 22 with version 1.9.2: an earlier update had saved gpt-3.5-turbo into sites’ settings as the default.
• Aliases. A request for o4-mini or gpt-4 fails as surely as one for the dated snapshot. Both names are on OpenAI’s list.
• Fine-tunes. A model you fine-tuned on GPT-3.5 Turbo or GPT-4.1 nano shuts down with its base model.
• Tools set to your API key. A script, automation or coding agent that sends o3-mini or o4-mini to the OpenAI API with your key gets the same failure as your own code.

To find them, search your code, configuration files and environment variables for each name in the table, and check settings stored in a database, where a plugin’s default can live. For the next round, OpenAI’s List models endpoint, GET /v1/models, returns an optional shutdown_date for each model, “as a YYYY-MM-DD string, or null if no shutdown date has been announced.”

What to switch to

OpenAI’s substitutes date from April, when GPT-5.6 was its newest family. It has since released GPT-6, and its October notices name GPT-6 models as replacements. Both families are on the API, each with a context window of 1.05 million tokens:

Model	Input / output, per million tokens
GPT-5.6 Sol	$4 / $20, promotional, at least through November 21, 2026
GPT-5.6 Terra	$2 / $12
GPT-5.6 Luna	$0.20 / $1.20
GPT-6 Astra	$10 / $50
GPT-6.1 Sol	$2 / $10
GPT-6 Luna	$0.10 / $0.50

Prices are from OpenAI’s models page and each model’s own page. What the move costs depends on where you start:

• From gpt-3.5-turbo, at $0.50 in and $1.50 out, the listed substitute, GPT-5.6 Terra, costs four times as much for input and eight times as much for output. GPT-6 Luna costs less than GPT-3.5 Turbo did.
• From o3-mini or o4-mini, at $1.10 and $4.40, both substitutes cost more per token. GPT-6.1 Sol matches Terra’s input price and costs less for output, and GPT-6 Luna costs less than either old model did.
• From gpt-4, at $30 and $60, or o1, at $15 and $60, every model in the table costs less per token.

The request changes too, especially from GPT-3.5 or GPT-4. OpenAI’s GPT-6 migration notes say to remove temperature, top_p and top_logprobs whenever reasoning effort isn’t none. GPT-6 Astra and GPT-6.1 Sol don’t support none; GPT-6 Sol and GPT-6 Luna do. Tool calling needs the Responses API, except that GPT-6 Sol and Luna support function calling in Chat Completions with reasoning_effort: "none". By those rules, an old Chat Completions request with a temperature and function calls moves most directly to GPT-6 Luna with reasoning effort none. Test it on real requests before you rely on it.

OpenAI says Codex can make these changes for you with its OpenAI Docs skill: $openai-docs migrate this project to the GPT-6 model family, or GPT-5.6 model family for the listed substitutes. Other coding agents can download the same skill. Review what it changes like any other diff.

If you can’t switch today

OpenAI leaves one door open: “In some cases, developers may be able to provision dedicated capacity for continued access after a model’s shutdown date,” through its sales team. On Azure, the reasoning models have until November 19.

For everyone else, the next dates are December 11 for o3 and the first GPT-5 models, and January 14, 2027 for GPT-4o, GPT-4o mini, GPT-4.1 and GPT-4.1 mini. Moving those requests in the same change saves doing it twice.