GPT-4o Returns as Default Model for All Paying ChatGPT Users; Altman Vows Advance Notice for Any Future Changes

OpenAI has reinstated GPT-4o as a default option for all paying users, including subscribers to ChatGPT Plus, Pro, Team, Enterprise, and Edu tiers. Previously accessible only by enabling a "show legacy models" setting, the language model that powered ChatGPT prior to last week's GPT-5 launch is now readily available once again.

At the same time, paying subscribers will also see a new “Show additional models” setting enabled by default, restoring access to GPT-4.1, o3, and o4-mini—the latter two being reasoning-focused models.

OpenAI CEO and co-founder Sam Altman announced the shift on X just minutes ago, promising that if GPT-4o is ever removed in the future, the company will provide "plenty of notice."

Updates to ChatGPT:

You can now choose between “Auto”, “Fast”, and “Thinking” for GPT-5. Most users will want Auto, but the additional control will be useful for some people.

Rate limits are now 3,000 messages/week with GPT-5 Thinking, and then extra capacity on GPT-5 Thinking…

— Sam Altman (@sama) August 13, 2025

These models appear in the model selection menu at the top of the ChatGPT session screen, both on the web and across mobile and other applications.

This move comes after a rocky first week for GPT-5, which launched on August 7 in four variants—regular, mini, nano, and pro—some of which included optional “thinking” modes tailored for longer, reasoning-heavy tasks.

As VentureBeat previously reported, GPT-5’s debut received mixed feedback and encountered technical issues, including a malfunctioning “autoswitcher” that misdirected user prompts, inconsistent performance relative to GPT-4o, and user frustration over the abrupt removal of older models.

Altman's recent update introduces additional controls to the ChatGPT interface, giving users the ability to select between “Auto,” “Fast,” and “Thinking” modes when using GPT-5.

The “Thinking” mode, which supports a 196,000-token context window, is now subject to a 3,000-message weekly limit for paying users. Once this cap is reached, subscribers may continue using the scaled-back “GPT-5 Thinking mini” mode. Altman indicated these limits could be adjusted based on usage patterns.

Notably, GPT-4.5 remains available exclusively to Pro users, reflecting its high GPU cost.

Altman also suggested a forthcoming personality adjustment for GPT-5, aiming for a tone that feels “warmer” than the current default while avoiding the polarizing nature of GPT-4o.

The company is exploring per-user customization as a long-term approach—a potential solution to the strong emotional bonds some users have developed with earlier models.

These updates are expected to ease frustration among users who were unsettled by the sudden switch to GPT-5 and the removal of older OpenAI models, though they may also sustain the intense attachments some individuals have formed with specific AI personalities.