After a difficult launch of GPT-5, GPT-4o is now once again accessible for paying ChatGPT subscribers. OpenAI CEO Sam Altman guaranteed users they would be notified prior to the subsequent removal of GPT-4o.

On Tuesday, Altman shared on X that legacy models, such as GPT-4o, GPT-4.1, o3, and o4-mini, are available in the dropdown menu for ChatGPT Plus and Pro subscribers. The 4o model can currently be located under “Legacy Models.”

Moreover, paying users can activate “Show additional models” in Settings to select from GPT-5 Thinking mini, GPT-4.1, o3, and o4-mini. GPT-4.5 is exclusively available to ChatGPT Pro subscribers due to its substantial GPU costs, as noted by Altman.

OpenAI brought back these models following the unexpected replacement caused by the GPT-5 launch, which led to dissatisfaction among users. GPT-5 was designed to function as a unified model that automatically adjusts computational resources based on the complexity of prompts. Previously, users had the option to choose models tailored to task complexity, with reasoning models like o3 and o4-mini appropriate for intricate prompts, and conventional models like GPT-4o for simpler dialogues.

The introduction of GPT-5 resulted in user frustration when favored models vanished. Users voiced their disappointment with GPT-5 for missing the warmth and character of GPT-4o. The shift was further complicated by the malfunction of ChatGPT’s “autoswitcher,” which impacted GPT-5’s effectiveness. OpenAI swiftly reacted by reinstating GPT-4o for paying subscribers.

Users now have the option to select “Auto” for GPT-5, which identifies when to utilize reasoning or “thinking,” or to manually choose “Fast” or “Thinking.” Altman addressed user worries regarding the loss of access to GPT-4o without prior notice, ensuring that there would be sufficient warning if it were to be deprecated. He also highlighted updates to GPT-5’s character to enhance its warmth and stressed the importance of greater user customization of model personalities.

Disclosure: Ziff Davis, the parent company of Mashable, initiated a lawsuit against OpenAI in April, alleging copyright violations in the training and operation of its AI systems.