GPT-4.5 Is Being Removed From ChatGPT on June 27 — Here’s What Changes and What Doesn’t

OpenAI is retiring its last GPT-4 model from ChatGPT in five days. If you’re a regular ChatGPT user, you probably don’t need to do anything. If you’re a developer or business running API integrations, you also don’t need to do anything — yet. Here’s the full picture.

What’s Actually Happening

On June 27, 2026, OpenAI is removing GPT-4.5 from ChatGPT. The model will no longer appear in the model picker for paid users, and any conversations that were using it will automatically continue on GPT-5.5.

Independent coverage of the GPT-4.5 and o3 retirements from GenAI Skool, published June 2026 — not a Wangdoo production.

Two things are important to state clearly upfront, because most coverage of this has muddled them:

• This only affects ChatGPT — not the API. If you use GPT-4.5 via OpenAI’s API, nothing changes on June 27. API access is unaffected, and OpenAI has not announced an API retirement date for GPT-4.5.
• GPT-4.5 was only available to paid users anyway. Free ChatGPT users never had access to GPT-4.5 in the model picker. This retirement only affects Plus, Pro, and Team plan subscribers who had explicitly selected it.

What GPT-4.5 Actually Was

GPT-4.5 launched on February 27, 2025, described by OpenAI in its own announcement as “our largest and best model for chat yet” — a model that scaled up pre-training rather than reasoning, deliberately positioned as a different kind of capability from the o-series reasoning models. It launched as a research preview for Pro plan users first, before rolling out to Plus, Teams, and Enterprise. OpenAI specifically highlighted its “EQ” — its ability to understand nuance, interpret subtle cues, and produce more natural, human-feeling conversation — alongside broader world knowledge and reduced hallucinations compared to GPT-4o.

That positioning matters for understanding the user reaction. GPT-4.5 was not designed to be the fastest or cheapest model, and it wasn’t the strongest at structured reasoning tasks — that was o3’s territory. What it offered was depth: more knowledge, warmer conversation, stronger aesthetic intuition in writing. Writers and researchers were its most consistent advocates, and that is precisely the audience that now feels most displaced by its removal.

It lasted approximately 16 months in ChatGPT — not a short lifecycle by recent standards, but GPT-4.5’s retirement still feels abrupt to its users because it never got the wider cultural moment that GPT-4o did. GPT-4o attracted genuine user attachment, sparked backlash when OpenAI tried to retire it in August 2025, and was brought back before being finally removed in February 2026. GPT-4.5 leaves more quietly. It is also the last remaining GPT-4 family model in ChatGPT. Once it goes on June 27, no GPT-4 model will be accessible inside the consumer product at all.

That reaction is representative of a specific segment of power users who valued GPT-4.5 not for benchmarks, but for the particular quality of its prose and the breadth of its responses. Whether GPT-5.5 matches it on those specific qualities is something users will find out on June 27 — OpenAI says it does; some users disagree.

What This Means Depending on How You Use ChatGPT

What Replaces It — and Whether It’s Actually Equivalent

Inside ChatGPT, GPT-5.5 is the direct replacement. OpenAI released GPT-5.5 on April 23, 2026, and has since updated it with specific improvements to response style — shorter, less bullet-heavy answers, more natural pacing, and better conversational flow. The most recent GPT-5.5 Instant update was explicitly targeted at the feedback OpenAI has consistently received since the GPT-4o retirement: that GPT-5.x models feel less warm and more formulaic than what came before. The broader pressure on OpenAI to keep ChatGPT feeling best-in-class is real — Apple’s decision to power its next-generation Siri with Google’s Gemini rather than OpenAI’s models is a reminder that the consumer AI landscape is genuinely competitive now, and ChatGPT’s model experience is one of the few levers OpenAI directly controls.

On most objective benchmarks, GPT-5.5 outperforms GPT-4.5 across coding, reasoning, and knowledge tasks. OpenAI’s own data showed GPT-4.5 had moved to “limited usage” before the retirement announcement, which broadly reflects how completely the GPT-5.x generation has taken over for general use.

The honest caveat remains the same as with every previous model retirement: benchmarks don’t capture writing voice, stylistic preference, or conversational feel. GPT-4.5’s “EQ” positioning was about something genuinely different from benchmark performance — a specific quality of response that some users experienced as more considered, more natural, and less performatively helpful. Whether GPT-5.5 has caught up on that dimension is something each user will need to judge for themselves. The best way to know is to use both before June 27 while you still can.

The Bigger Pattern Worth Knowing

OpenAI’s model retirement cadence is accelerating. The company is moving toward a consumer product that runs exclusively on its current generation — a tightly curated GPT-5.x experience — while the API maintains broader compatibility as a separate track. The practical implication for anyone building on ChatGPT or OpenAI’s API is that model availability in the consumer product and API availability are increasingly decoupled.

The asymmetric sunset windows are telling: GPT-4.5 got 30 days while o3 gets 90 days. OpenAI’s own explanation is that models with “limited usage” are retired on shorter timelines. The longer window for o3 reflects that more users had built workflows around its specific reasoning capabilities — o3 was not just a general-purpose model but one many users selected deliberately for complex analytical tasks.

Frequently Asked Questions

My Take — Mr Wangdoo

The users upset about GPT-4.5 leaving aren’t wrong to be upset — they’re describing a real thing. GPT-4.5 had a particular quality to its prose that a specific group of writers and researchers genuinely valued, and “the new model is better on benchmarks” doesn’t fully address that. OpenAI made the same mistake with GPT-4o, tried to retire it, got significant pushback, brought it back, and then retired it again once GPT-5.x had actually addressed the concerns. The company listened the first time round. Whether GPT-5.5 has genuinely resolved what GPT-4.5 users valued, or whether this retirement will follow the same pattern of complaints, is something the next few weeks will answer.

The more important pattern here isn’t about one model — it’s about how OpenAI is structuring its products going forward. A consumer-facing ChatGPT that runs on one consolidated generation, and an API that maintains wider compatibility for developers and enterprises, are two fundamentally different products now. If you’re a regular ChatGPT user, you’re increasingly on OpenAI’s curated experience rather than a menu of options. If you’re a developer, the API track is where your control lives. Those are different relationships with the same company, and it’s worth being clear-eyed about which one you have.

Sources

• Introducing GPT-4.5 — launch announcement, positioning, and capability description — OpenAI
• ChatGPT Release Notes — GPT-4.5 and o3 retirement announcement — OpenAI Help Centre (primary source)
• Model Release Notes — retirement cadence and API status confirmed — OpenAI Help Centre
• Retiring GPT-4o, GPT-4.1, and other models — context on OpenAI’s retirement policy and GPT-4o backlash — OpenAI Blog
• User reaction to GPT-4.5 retirement (“best writing model”) — X / Threads, widely reported June 2026