OpenAI regularly makes adjustments to the ChatGPT model selection. Older models are removed, and new variants are introduced. This usually happens without much fanfare. This time, however, the announcement is more extensive, as several well-known models are affected, and a previous promise from the company plays a central role.
Since the introduction of more powerful language models, OpenAI has been continuously modifying the structure of ChatGPT. This involves not only technical advancements but also usage frequency, costs, and user behavior. Models with clearly recognizable personalities, in particular, have demonstrated in the past that OpenAI's decisions can elicit strong reactions.
Which models will be removed from ChatGPT
On February 13th, several models will be removed from ChatGPT's model picker. According to OpenAI, these are the following variants:
- GPT-4o
- GPT-4.1
- GPT-4.1 mini
- OpenAI o4-mini
These models will no longer be directly selectable in ChatGPT. It's important to note that this change applies exclusively to ChatGPT. OpenAI explicitly states that there are no changes to the API "at this time." Developers can still use these models via the API.
The story of GPT-4o
The final shutdown of GPT-4o is receiving particular attention. When OpenAI released GPT-5 in August of last year, GPT-4o was removed from the model picker without much warning. This decision immediately triggered a strong backlash.
Many users had grown accustomed to the tone and behavior of GPT-4o. The model was often perceived as warmer, friendlier, and more personal than GPT-5, whose responses were considered more direct and matter-of-fact. Its sudden removal prompted OpenAI to make GPT-4o available again on short notice.
Reactions and ethical debates
The return of GPT-4o went beyond a mere product decision. It sparked intense discussions about people's emotional attachment to large language models. These discussions centered on questions of responsibility, expectations, and how much personality should be a part of an AI system.
In this context, Sam Altman, CEO of OpenAI, also commented. He stated that GPT-4o is once again available by default for all paying users. At the same time, he promised that any potential shutdown would be announced in advance.
Adjustments to ChatGPT and new options
In the same statement, Altman referred to new settings within ChatGPT. Paying users received the option "Show additional models," which allows them to activate further models such as o3, GPT-4.1, or GPT-5 Thinking mini. Model 4.5 remained reserved for Pro users, as it requires significant GPU resources.
In parallel, OpenAI announced it was working on the personality of GPT-5. The goal was a warmer tone that would appear less distant, without adopting the mannerisms of GPT-4o, which many users found too intrusive. A key insight from this phase was that model personalities need to be more tailored to individual preferences.
Why GPT-4o is now being switched off
OpenAI has since implemented these adjustments. Users can now choose from seven different personalities and also add their own instructions. Despite this increased personalization, the use of GPT-4o has decreased significantly.
According to OpenAI, only about 0.1 percent of users interact with this model daily. Based on this, it was decided to permanently remove GPT-4o from ChatGPT.
- ChatGPT Health put to the test: When AI misreads heart data
- ChatGPT Atlas: New update improves the AI browser
- OpenAI is specifically preparing for its first hardware launch in 2026
- ChatGPT introduces affordable subscription and tests advertising
Data-driven decisions at OpenAI
The current This decision demonstrates how strongly OpenAI aligns its product strategy with real-world usage data. At the same time, the GPT-4o case illustrates the close connection between technical development and user perception. While ChatGPT continues to be simplified and modernized, older models will remain available, at least for developers, for the time being. OpenAI is thus continuing its strategy of streamlining ChatGPT without directly restricting existing APIs. (Image: Shutterstock / Harry Howitt)
- Apple TV announces release date for Sugar Season 2
- Tim Cook believes Apple's best work is yet to come
- Apple stock in focus: JP Morgan raises price target to $325
- N2See for macOS: Control over windows on external displays
- Apple was surprised by the demand for AirPods Pro 3
- Apple explains its strategy in the face of rising RAM and SSD prices
- Tim Cook: Apple remains committed to data privacy despite Google partnership
- Apple reports 2.5 billion active devices worldwide
- Apple Q1 2026: Record revenue and strong profit growth
- Apple TV celebrates F1 as the most-watched film of all time
- Apple confirms acquisition of Q.ai for $2 billion
- The iPhone 16 will top the smartphone charts in 2025
- Apple strengthens design team with Halide co-founder
- Apple secures the rights to the Cosmere universe
- iOS 26.3 Beta 3 Warning: These Continuity features will be unavailable
- Patreon and Apple: New deadline once again draws criticism
- Apple Watch shows benefits in study on AFib aftercare
- Apple loses top position at TSMC due to AI boom
- AirTag 2: Teardown reveals all internal innovations
- Apple poised for a strong quarter: iPhone 17 and AI in focus
- Apple Creator Studio: new creative subscription now available
