OpenAI Is Retiring GPT-4.5: What ChatGPT Users Need to Know

GPT-4.5 launched in February 2026 as the most capable model OpenAI had shipped. It's gone from ChatGPT by June 27.

OpenAI announced this week that GPT-4.5 will be retired from the ChatGPT interface on June 27, 2026, alongside GPT-4o and GPT-4.1. OpenAI o3 gets its own August deadline. The company is consolidating a sprawling model menu down to the GPT-5 generation, and the GPT-4 era is the casualty.

This is faster than most users expected. Four months from debut to end-of-life is short even by the accelerated standards of the AI industry. The speed reflects a specific calculation: GPT-5 already does everything GPT-4.5 does, generally better and at lower cost per token. According to OpenAI's retirement announcement, the company made the decision after GPT-5 exceeded GPT-4.5's capabilities across standard benchmarks.

What the Retirement Actually Covers

The June 27 date applies to ChatGPT only. GPT-4.5 through the API is on a separate deprecation timeline, and OpenAI has not announced equivalent end dates for API access. If you're a developer whose application calls GPT-4.5 via the API, your situation is different and more complex than the ChatGPT retirement.

For ChatGPT users, the retirement means the manual model selection option disappears from settings. If you've been picking GPT-4.5 from the dropdown for specific tasks, ChatGPT will route those conversations to GPT-5 automatically after June 27.

The post-retirement model menu becomes cleaner: GPT-5 as the general-purpose default, GPT-5.3-Codex for code-intensive work, and o4-mini for reasoning-heavy tasks. That's the working roster for the near term.

Who Actually Loses Anything Here

The honest answer is: very few people.

GPT-4.5 was only available on paid plans. Free-tier users never had access to it, so they won't notice its absence. ChatGPT Plus and Pro subscribers who had access will be automatically migrated to GPT-5, which outperforms GPT-4.5 across writing quality, instruction-following, and factual accuracy. BleepingComputer's coverage of the announcement confirms GPT-5 is faster and cheaper per token as well.

There are narrow cases where users who calibrated prompts to GPT-4.5's specific behavior patterns may notice output style differences. GPT-5 is not GPT-4.5 with more capability — it has different response tendencies. Most users will find GPT-5's defaults preferable, but some will need to adjust workflows built around GPT-4.5's particular tone and formatting habits.

Three Models Worth Knowing Going Forward

The post-retirement ChatGPT lineup is more navigable than what's been available for the past several months.

GPT-5 handles the majority of what ChatGPT users actually do: writing, summarization, research assistance, brainstorming. It's the correct default for most tasks, and there's no strong reason to override it unless you have a specific use case.

For software development work, GPT-5.3-Codex is worth the explicit switch. Released in May 2026, it outperforms base GPT-5 on debugging, code explanation, and multi-file reasoning. If programming is a significant part of how you use ChatGPT, choosing Codex deliberately matters. The Best AI Chatbots for Coding in 2026 covers this alongside alternatives like GitHub Copilot and Cursor.

For math, formal logic, and multi-step reasoning, o4-mini remains the most reliable ChatGPT option. It's slower, which is a real tradeoff. But for tasks where accuracy on a chain of logical steps matters more than response speed, the latency is worth accepting.

What Model Retirements Reveal About AI Assistants

The GPT-4.5 retirement isn't unusual in isolation. It's the fourth significant ChatGPT model sunset in 14 months. OpenAI retired GPT-4.5 faster than it retired GPT-4 Turbo, and faster than GPT-4o is going. The cadence is accelerating.

This matters for anyone building habits or workflows around a specific model's behavior. Faster model cycling means faster access to capability improvements. It also means less time to learn what a model does well and less stability for tools built on top of it.

Anthropic's Claude 3.7 Sonnet has operated on a longer stability window — the company typically keeps models active for six or more months before cycling. Google Gemini 2.5 Pro has maintained a similarly slower cadence. Neither approach is inherently superior to OpenAI's. But if workflow stability weighs more than capability recency for your work, the retirement calendar is now a legitimate factor in which AI assistant you choose as your primary tool.

OpenAI retires models quickly because newer versions are genuinely better. For most ChatGPT users, GPT-5 will deliver a noticeably better experience than GPT-4.5 did. The broader AI assistant landscape is worth considering too — Apple is making significant moves ahead of WWDC on June 8, and the Best Free AI Chatbots of 2026 offers a wider comparison if you're evaluating alternatives.

---

The About.chat newsletter covers AI changes that actually affect how you work — no hype, just what matters. Subscribe here to get it every week.