GPT-5.2 is rolling out — and GPT-4o disappears from ChatGPT on February 13

A lot of people only notice platform changes when something is gone: OpenAI is simplifying ChatGPT. On February 13, 2026, several models disappear from ChatGPT’s model picker — including GPT-4o. At the same time, GPT-5.2 has been rolling out for a while and is meant to become the default experience for most users.

Key point: this is not automatically the same thing as an API shutdown. ChatGPT and the API have different lifecycles.

What changes on February 13 in ChatGPT?

According to OpenAI, ChatGPT will remove these models on February 13, 2026:

• GPT-4o

• GPT-4.1

• GPT-4.1 mini

• OpenAI o4-mini

This happens alongside a previously announced retirement of GPT-5 (Instant, Thinking, and Pro) in ChatGPT.

In practice:

• The model picker gets smaller.

• Workflows that “sound right on GPT-4o” will get pushed onto newer models.

What (still) doesn’t change in the API

OpenAI explicitly says: there are no API changes at this time tied to the ChatGPT retirement.

That’s reassuring, but incomplete. In the API world, OpenAI typically deprecates specific model snapshots, not a brand name.

For example, OpenAI’s official deprecations page states that the chatgpt-4o-latest snapshot was deprecated earlier and will be removed from the API on February 17, 2026.

Why OpenAI is doing this: GPT-5.2 as the new standard

GPT-5.2 is framed as a model family for “professional work and long-running agents.” Ignore the slogans; the practical drivers are straightforward:

• Fewer models to operate in parallel

• Less confusion and fewer edge-case support issues

• A single default that behaves more reliably for most people

Your job is not to care about names — it’s to ensure your outputs stay consistent.

What you should do now (without overreacting)

1) Do a quick inventory

Write down:

• Which prompts, workflows, or custom GPTs depend on GPT-4o

• Where style is critical (support replies, public-facing tone, legal-ish text)

2) Test the switch to GPT-5.2 on purpose

Run a small regression-style review with 10–20 representative examples:

• “Normal” cases

• “Nasty” cases (long inputs, delicate wording, multi-step instructions)

If the output shifts, adjust the prompt first before rewriting your product.

3) Add basic model resilience (API)

If you use the API:

• Don’t hard-depend on “latest” snapshots.

• Keep a simple fallback chain (e.g., primary GPT-5.2, secondary a stable alternative).

• Log model name/version per request so debugging stays possible.

4) Communicate early internally

If teams have strong expectations (“GPT-4o feels warmer”), be explicit:

• ChatGPT’s model picker changes on Feb 13.

• API timelines may differ.

• You need a short re-validation pass.

Common misconceptions

“GPT-4o is dead everywhere on Feb 13.”

No. Feb 13 is the ChatGPT UI cutoff. API removals target specific snapshots and can happen later.

“GPT-5.2 is automatically better at everything.”

No. It’s often stronger, but some tasks are sensitive to prompting and tone. Test.

“There’s nothing I can do.”

There is: prompt tuning, regression tests, and a minimal fallback strategy dramatically reduce risk.

Bottom line

OpenAI is moving ChatGPT toward GPT-5.2 and removing older models like GPT-4o from the ChatGPT interface on February 13, 2026. If you rely on the API, the dates you must care about live in the deprecation notes for specific model snapshots.

Do a small inventory and a short verification run against GPT-5.2 now, and the transition is usually painless — while your system becomes bigger and more robust instead of being glued to a single model name.

Sources:

• OpenAI: Retiring GPT-4o, GPT-4.1, GPT-4.1 mini, and OpenAI o4-mini in ChatGPT (Feb 13, 2026) https://openai.com/index/retiring-gpt-4o-and-older-models/

• OpenAI API Docs: Deprecations (includes chatgpt-4o-latest removal on Feb 17, 2026) https://platform.openai.com/docs/deprecations

• OpenAI: Introducing GPT-5.2 https://openai.com/index/introducing-gpt-5-2/