OpenAI has quietly retired GPT-4.5, the last remaining GPT-4 model from ChatGPT, closing the chapter on the AI family that transformed generative AI from a curiosity into a mainstream tool. The company announced the retirement in a brief release note, stating it was retiring older models with limited usage to better serve newer, more capable versions. GPT-4.5 will disappear on June 27, 2026, following a 30-day sunset period, while the reasoning model o3 will be retired on August 26, 2026, after a 90-day sunset period.

Why Does the Retirement of GPT-4 Matter?

GPT-4 and its variants, particularly GPT-4o, were the models that convinced millions of people that artificial intelligence had genuinely arrived. These models could write essays, pass exams, code applications, and analyze images, making AI feel intelligent rather than gimmicky for the first time. The retirement of GPT-4.5 signals that no GPT-4 models remain in ChatGPT, marking a symbolic end to a remarkable chapter in AI history.

The move is not without controversy. When OpenAI previously retired GPT-4o in favor of the newer ChatGPT-5, the backlash was significant enough that the company brought it back temporarily before ultimately removing it once ChatGPT-5 had improved sufficiently. Some users have already expressed frustration on social media about losing GPT-4.5. One user commented that "4.5 is the best writing model" and that the newer GPT-5 series "still doesn't match what those two had," suggesting that capability improvements don't always translate to user satisfaction.

How Does AI Model Retirement Work?

• Sunset Period: OpenAI provides users with advance notice and a grace period before fully removing models, allowing developers and users to transition to newer alternatives.
• Availability Restrictions: Both o3 and GPT-4.5 are currently only accessible to paid ChatGPT users through the model settings menu, limiting their reach before retirement.
• Replacement Strategy: OpenAI retires older models to consolidate its offerings and direct users toward newer, more capable versions like GPT-5.5 and the latest reasoning models.

The retirement process reflects a broader industry pattern where AI companies continuously iterate and upgrade their offerings. However, the emotional attachment users have developed to specific models complicates this technical necessity. AI progress doesn't always feel like a straight line, even when benchmarks improve and capabilities expand.

What's Driving User Nostalgia for Older Models?

Interestingly, the backlash to GPT-4.5's retirement reveals something unexpected about how people relate to AI. For the first time in computing history, users aren't just nostalgic for old software; they're nostalgic for old personalities. The personality and feel of a chatbot can have more influence on user preference than almost anything else, according to reporting on the topic. New models may arrive with better benchmarks, faster responses, and more capabilities, yet they can also lose some of the qualities that made people connect with earlier versions.

This phenomenon suggests that AI adoption isn't purely rational. Users develop preferences based on how a model communicates, responds to nuance, and handles creative tasks, not just raw performance metrics. GPT-4 and its offshoots transformed AI from a fascinating curiosity into something people genuinely used, relied on, argued about, and in some cases even formed attachments to. The retirement of GPT-4.5 deserved more than a passing mention in a product release note, given its cultural significance.

OpenAI is betting that few people will miss GPT-4.5 now that successors like GPT-5.5 have taken over. Yet the previous backlash surrounding GPT-4o's retirement suggests the company may face similar pushback. As AI models become more integrated into people's workflows and creative processes, the transition between versions may prove more contentious than simple software updates of the past.