GPT-5 update: Altman explains three new changes to ChatGPT, after backlash

When OpenAI launched GPT-5 earlier this month, it was meant to be a milestone moment, a demonstration of how far artificial intelligence had come in reasoning ability, adaptability, and real-world usefulness. Instead, it triggered an unexpectedly strong backlash.

Survey

Long-time ChatGPT users, especially those on paid subscriptions, complained about losing access to older models, running into stricter usage caps, and feeling that the new system behaved differently from what they were used to. Some even described GPT-5 as “too serious” or “less creative” compared to GPT-4o, which it replaced as the default model.

Add As A Trusted Source For Google.

Add as a preferred source on Google

Now, CEO Sam Altman has responded with three significant changes to ChatGPT’s experience, aimed at giving users back more control and flexibility.

Also read: Altman vs Musk battle turns funny with ChatGPT vs Grok fight: Whom to trust?

New response modes: Auto, Fast, and Thinking

The first change is the introduction of three selectable response modes, a way for users to choose not just what ChatGPT says, but how it thinks and delivers answers.

• Auto: The default mode, designed to automatically balance speed and depth depending on the query. For example, a quick fact-check will get a faster reply, while a complex research request will get more detailed output.
• Fast: Prioritizes speed above all else. Useful for rapid back-and-forth conversations, quick brainstorming, or when you’re in a hurry.
• Thinking: The “slow cooking” option. GPT-5 takes extra time to process, aiming for deeper reasoning, more nuance, and more carefully structured answers.
  

Altman noted that the Thinking mode comes with limits, Plus subscribers get around 3,000 messages per week, after which they can use a lighter “Thinking mini” model for extended reasoning without the same computational cost.

This system is meant to let users decide how much time and processing power they want ChatGPT to spend on each task. In Altman’s words: “Sometimes you want speed, sometimes you want depth. Now you can choose.”

GPT-4o is back (for paid users)

The second change addresses one of the loudest complaints from the GPT-5 rollout, the removal of GPT-4o, a popular multimodal model that many users had grown attached to. GPT-4o is now restored for Plus and Pro subscribers via a new “Show additional models” setting.

Also read: Sam Altman admits killing GPT-4o after GPT-5 launch was a mistake: Here’s why

This toggle also allows access to other non-default models, including o3, 4.1, and GPT-5 Thinking mini. GPT-4.5 remains exclusive to Pro subscribers. Many had described GPT-4o as friendlier, more intuitive, or simply better suited to certain creative tasks. Workflows built around GPT-4o’s specific style suddenly broke when GPT-5 became the default. Bringing it back, even if only for paying users, is a recognition that AI upgrades can’t ignore user preference. “We underestimated how much attachment there was to GPT-4o,” Altman admitted.

Higher rate limits for Plus subscribers

The third update is all about capacity. GPT-5’s initial rollout came with stricter usage limits than some expected, leading to frustration when users hit their caps mid-project. Now, rate limits for Plus subscribers have been doubled, particularly for GPT-5 in Thinking mode.

Altman said all model caps are now higher than they were before GPT-5 launched, giving power users more breathing room before they run into restrictions.

These changes reveal a clear lesson: AI progress isn’t just about raw capability, it’s about how people experience and trust the product. Even if GPT-5 is more advanced on paper, sudden changes to default models, speed, and limits can alienate loyal users.

By introducing response modes, restoring GPT-4o for those who want it, and expanding limits, OpenAI is moving toward a more customisable ChatGPT. It’s a pivot from “one size fits all” to “you’re in control.”

The GPT-5 backlash, and OpenAI’s rapid adjustments, show that AI companies must balance innovation with user familiarity. As Altman put it in his closing post on the matter: “We’re learning in real time, and we’ll keep adjusting.”

Also read: Early reactions to ChatGPT-5 are all bad: What went wrong for OpenAI?

Follow Us

Vyom Ramani

A journalist with a soft spot for tech, games, and things that go beep. While waiting for a delayed metro or rebooting his brain, you’ll find him solving Rubik’s Cubes, bingeing F1, or hunting for the next great snack. View Full Profile