ChatGPT: OpenAI terminates GPT-4o, GPT-4.1, GPT-4.1 mini and o4-mini in the application

At the start of 2026, OpenAI confirmed a decision that marks the end of an era in the evolution of ChatGPT: several historical models, including GPT-4oGPT-4.1, GPT-4.1 mini and OpenAI o4-mini will be removed from the ChatGPT interface starting February 13, 2026. This announcement, published in January 2026 on the company's official blog, comes as part of a strengthening of the standard model focused on the latest versions, notably GPT-5.2, which is becoming the default model for the majority of users.

Why OpenAI is removing these old models

OpenAI explains that the decision was not taken lightly: these models are being gradually phased out because of the use of newer models like GPT-5.1 or GPT-5.2 has far surpassed that of previous versions, and most users hardly use it anymore. According to the company itself, only 0,1% of users continued to choose GPT-4o every day before this decision.

By simplifying its portfolio of accessible models in ChatGPT, OpenAI aims to focus its development efforts on versions that offer the best overall performance, increased customization and more advanced features, while facilitating the user experience by reducing the fragmentation of available choices.

What this means for users

From February 13, 2026:

• The GPT-4o, GPT-4.1, GPT-4.1 mini and OpenAI o4-mini models will no longer be available in the ChatGPT interface for direct selection.

• New conversations and replies generated after this date will use GPT-5.2 by default, a newer and more powerful model.

• Existing conversations using older models will continue to appear in the history, but soon they will automatically switch to GPT-5.2 if you relaunch them.

• Developers who still use these models via the OpenAI API will be able to retain their access for the time being, even after the option is removed from ChatGPT.

In other words, your cats don't disappear, but they will no longer use the old patterns to generate new responses.

Why this transition now?

This decision can be understood as a natural phase of technological evolution. In the world of generative artificial intelligence, models are constantly evolving: new versions correct limitations, improve reasoning, broaden contextual understanding, and offer greater personalization of responses.

For example, GPT-5.2 incorporates user feedback that allows it to reproduce — or even surpass — some of the strengths of older models, while providing increased capabilities in the following areas: logical performance, natural language processing, multimodal understanding, and tone and style control.

This strategy also aligns with a logic of resource rationalization: maintaining multiple generations of models is costly in terms of infrastructure, monitoring, and maintenance. By retiring the least used versions, OpenAI can concentrate its energy on the dominant models, which are more efficient and more useful to the majority.

Community reactions

The announcement sparked mixed reactions within the user community. ChatGPT :

• Frustration among some users, particularly those who appreciated the conversational nature and "warmth" of GPT-4o, which some felt offered a more human and engaging dialogue experience. During a similar attempt in 2025, users protested online, stating they had developed a personal attachment to that style of AI.

• Acceleration towards newer models: others believe that the rise of GPT-5.1 and GPT-5.2 makes this transition logical, as the newer versions offer better overall quality and greater adaptability to complex tasks.

• Developers reassured: for those who still rely on old models in their applications, the ability to continue accessing them via the API during a transition period is a useful measure.

Smooth transition for advanced users

For certain professional or educational users (for example, in the Enterprise or Education sectors), OpenAI still offers an option to access legacy models during a transition period. This means that, under certain conditions, it is possible to activate the old models to maintain existing workflows or adapt internal systems before their use in ChatGPT is definitively discontinued.

This allows organizations to gradually adapt their tools and processes, without being suddenly forced to change their model overnight.

What changes with GPT-5.2

The withdrawal of the old models does not mean a loss of quality: on the contrary, GPT-5.2 becomes the reference model, combining and improving most of the strengths of its predecessors.

Key features of GPT-5.2 include:

• ➤ an expanded contextual understanding, enabling longer and more complex discussions to be handled;

• ➤ improved capabilities in content creation, analysis and multimodal generation;

• ➤ a personalization of the tone and style of response.

Thus, even if some people miss the style of older models like GPT-4o, the general user experience tends to unify around more powerful, more versatile and easier-to-maintain versions.

What this means in France

In France, this update comes at a time when many companies, startups and educational institutions are turning to recent ChatGPT models to modernize their workflows, create content, automate administrative tasks or analyze data.

Harmonisation around GPT-5.2 also means greater consistency in linguistic results, including for the French language, and access to more advanced functions in translation, writing, professional assistance, or multimodal tools.

For French-speaking users, this translates to:

• a more stable availability of recent models;

• an overall improvement in the quality of responses;

• the gradual end of older versions that were no longer aligned with current needs.

In short

The retirement of GPT-4o, GPT-4.1, GPT-4.1 mini, and OpenAI o4-mini from ChatGPT on February 13, 2026, marks a significant milestone in the history of consumer artificial intelligence. While these models represented different phases in ChatGPT's evolution, the majority of users are now migrating to newer versions like GPT-5.2, which offer a richer, more customizable, and higher-performing experience. This transition, despite some mixed reactions, reflects OpenAI's commitment to focusing its efforts on models that meet the current and future needs of users.