OpenAI has turned ChatGPT model selection into a deadline calendar. GPT-5.2 models were removed from ChatGPT on June 12, 2026, and the company’s current release notes also put GPT-4.5 and OpenAI o3 on dated retirement paths for paid ChatGPT users.
The change matters because many people treat ChatGPT’s model picker as a stable set of named tools. It no longer works that way. OpenAI is moving users toward newer default models, simplifying the model picker, and keeping older models available only through temporary legacy access or separate API lifecycle rules.
According to OpenAI’s ChatGPT release notes, GPT-5.2 Instant, GPT-5.2 Thinking, and GPT-5.2 Pro are no longer available in ChatGPT as of June 12. Conversations that used GPT-5.2 will continue automatically on the corresponding GPT-5.5 model. OpenAI’s release notes describe that as part of a broader policy in which ChatGPT models generally remain available for 90 days after a successor is released.
The Next Dates to Watch
The next two ChatGPT model deadlines are more awkward because they affect models some paid users may have selected deliberately for specific work styles.
- GPT-4.5 retires from ChatGPT on June 27, 2026. OpenAI lists this as a 30-day sunset for a model available to paid users through model settings.
- OpenAI o3 retires from ChatGPT on August 26, 2026. OpenAI lists this as a 90-day sunset for the reasoning model in ChatGPT.
- The API is not changing because of those ChatGPT deadlines. OpenAI’s note says the o3 and GPT-4.5 retirements apply to ChatGPT only.
That last point is easy to miss. A ChatGPT model retirement does not automatically mean the same model disappears from developer tools on the same date. OpenAI’s separate API deprecations page has its own shutdown schedule, including different deadlines for model snapshots, image models, Agent Builder, Evals, fine-tuning availability, the Assistants API, and older GPT aliases.
What Users Should Check Now
For ordinary ChatGPT users, the practical risk is not that old conversations disappear. The bigger issue is that a workflow tuned around a particular model can behave differently after it is moved to a newer default. That can affect writing tone, reasoning depth, coding style, formatting, and how often ChatGPT decides to use tools such as search, files, or connectors.
If you have recurring work that depends on GPT-4.5, o3, or a recently retired GPT-5.2 model, open a few representative chats now and rerun the same prompts with the newer model options. Compare the answers for instruction-following, length, citations, code correctness, and whether the model still handles the edge cases you care about.
That comparison is especially important for people who use ChatGPT as part of a repeatable process: drafting client deliverables, reviewing code, generating reports, checking policies, preparing lessons, or working from a saved custom prompt. The safest migration test is not a general vibe check. It is a short set of real prompts that match the work you actually do.
Enterprise and Edu Admins Have a Separate Setting
OpenAI’s legacy model access guidance says Enterprise and Edu workspaces can expose select older models only when legacy model access is enabled in workspace settings. The listed legacy models include GPT-4o, GPT-4.1, GPT-4.1 mini, OpenAI o4-mini, OpenAI o3, o3 Pro, and GPT-4.5, subject to availability windows and usage limits.
For admins, the job is not simply deciding whether to leave a toggle on. Teams should identify which groups still rely on older models, why they rely on them, and whether the use case is better served by a newer ChatGPT model, an API integration, or a documented workflow change. A sales team using a saved GPT for proposal drafts has a different migration risk than an engineering team comparing reasoning models for code review.
ChatGPT and API Lifecycles Are Splitting
The most useful way to read OpenAI’s current model changes is as two overlapping lifecycle systems. ChatGPT is becoming a managed product experience, with simplified picker labels such as Instant, Medium, High, Extra High, Pro Standard, and Pro Extended. The API remains a more explicit engineering surface, with named model snapshots, shutdown dates, and recommended replacements.
That split is sensible, but it creates a documentation burden for anyone who uses both. A product manager may talk about “o3 in ChatGPT,” while an engineer may be tracking a dated API snapshot, a fine-tuned base model, or a replacement path in the deprecations table. Those are not always the same thing.
The cleanest habit is to track three fields for any OpenAI-dependent workflow: where the model is used, which exact model or picker option is selected, and what date its current access path changes. For ChatGPT users, that means checking release notes and workspace settings. For developers, it means checking the API deprecation table, not assuming the ChatGPT picker tells the whole story.
Bottom Line
GPT-5.2 is already gone from ChatGPT, GPT-4.5 leaves on June 27, and o3 leaves on August 26. Existing chats may continue on newer models, and the API has its own schedule, but the habit of anchoring work to a favorite ChatGPT model name is becoming fragile.
For anyone who depends on ChatGPT professionally, the next step is simple: test your important prompts against the replacement models before the deadline, document the model choice in repeatable workflows, and treat model availability as part of normal software and operations planning.
