OpenAI to phase out o3 and GPT-4.5 models, sparking backlash

OpenAI has confirmed the retirement of two models from ChatGPT; the company will phase out GPT-4.5 in June and the OpenAI o3 model in August of this year.

GPT-4.5 and o3 were only available to paid subscribers. Now, they’re sunsetting. This was announced directly on the OpenAI Help Center. The post cites limited usage among paid subscribers as the primary reason, stating that the move allows them to focus resources on newer systems like the GPT-5 series.

It’s worth noting that access to APIs remains unchanged, and it only affects direct ChatGPT usage.

In the Help Center, OpenAI has also mentioned the release of a new version of GPT-5.5 Instant. Responses are now easier to read, and they feel more natural in conversations. There’s also better pacing, and it tries to avoid overly lengthy answers.

Additionally, the Canvas feature is also being taken down on the GPT-5.5 series of models. Writing and coding tools will now be built directly into the regular chat. Paying users will have access to Canvas until o3 and GPT-4.5 are still available, but others won’t.

The company has given GPT-4.5 a 30-day sunset period, so it’ll continue to work as usual from May 28 to June 27. Meanwhile, the o3 model has a 90-day period, and you can use it without interruptions from May 28 to August 26.

People described o3 as a model with a distinctive personality and a natural flow of language, alleging that the newer GPT-5 variants aren’t the same. Meanwhile, GPT-4.5 was praised as a valuable tool for writing, since it doesn’t “sensationalize” sentences and make simple phrases unnecessarily complex. Users are fairly disappointed that these models are going away.

Communities on Reddit and X highlighted how OpenAI’s retirement of models that people were actively paying for doesn’t make sense. These paying users also argue that low usage statistics may have several other reasons, such as rate limits, and that lack of demand isn’t the only conclusion to draw from these statistics.

One key reason for the lack of usage is that the model is hidden behind a paywall in the first place. People are annoyed with OpenAI in general for prioritizing rapid iteration over preserving models that people have a connection with.

Broader conversations spoke about a shift towards controlled, corporate-safe AI models, which sacrifice personality. While OpenAI has introduced a “less robotic” GPT-5.5 Instant model, it remains to be seen whether it actually satisfies users as much as 4.5 and o3 did.

Additionally, an employee also shared information recently that the company is “making space for what’s coming.” This has led to widespread speculation about GPT-5.6 potentially dropping today.

This isn’t the first time the company has faced strong pushback from its user base. Earlier this year, the Keep4o movement gained significant traction as users protested the planned retirement of GPT-4o. More on that here. 

Around the same time, a fake screenshot claiming OpenAI was offering select users permanent access through a non-existent GPT-4o Legacy Program circulated online, which was called out as a scam. You can read about that here.

Disclaimer: The Featured image was generated with AI.

We stand out from the tech-media crowd because we break news stories; we mainly bring you stuff that you won’t find anywhere in the mainstream tech media. Our stories have been picked up by some of the world’s most popular websites and media outlets—more info is available here.