Why are OpenAI's legacy GPT models returning to ChatGPT? We may have answer

After criticism over removing older models, OpenAI now offers GPT-5 in multiple modes and restores access to popular legacy models like GPT-4o, GPT-4.1, and o3

OpenAI has returned with a long list of AI models in ChatGPT to choose from. OpenAI launched GPT-5 recently with the promise that it would remove the hassle for users to choose the model that they want to interact with, including GPT-4o. After facing backlash from users, OpenAI has introduced “Auto”, “Fast”, and “Thinking” settings for GPT-5 that all ChatGPT users can select from the model picker. Additionally, it has also added all other existing models GPT-4o, GPT-4.1, and o3 under the “Legacy models” banner.

What has happened

When OpenAI released GPT-5, it introduced GPT-5, GPT-5 Thinking, and GPT-5 Thinking Pro models. The company assured that OpenAI would automatically choose between these models depending on the query asked. It also removed all other models that were in use till the day before the release of GPT-5.

 

Many users lamented the change, as they had formed a bond with models like GPT-4o. Following this widespread backlash, OpenAI Chief Executive Officer Sam Altman announced that they will retain the GPT-4o model for paid users, and they can opt to use it instead of the newer GPT-5 models.

 

Now, Altman has not just introduced “Auto”, “Fast”, and “Thinking” settings for GPT-5 that all ChatGPT users can select from the model picker but, it has also retained access to several legacy models, including GPT-4o, GPT-4.1, and o3 — which were deprecated last week.

What does this mean

On top of OpenAI choosing between the modes automatically, users now have a list to manually choose the AI model that they want to interact with.

 

Additionally, for those using GPT-5, ChatGPT will offer the following modes:

• Auto: This mode lets ChatGPT decide the best mode for your query without manual selection. It balances speed and depth, switching between modes as needed.
• Fast: Focuses on delivering quick responses with minimal processing time. Best for short, straightforward questions or when speed matters most.
• Thinking: Takes more time to generate responses for greater depth and accuracy. Ideal for complex tasks that require reasoning, analysis, or detailed explanations.

For the near future, Altman wrote in his X post that the team is working on an update to GPT-5’s personality, which should feel warmer than the current personality but not as “annoying (to most users)” as GPT-4o.