When OpenAI launched GPT‑5 on August 7, 2025, the AI model arrived with ambitious promises: superior reasoning, advanced coding and writing capabilities, and more accurate knowledge compared to predecessors . Yet the rollout was far from smooth. Within days, users – especially those on the ChatGPT Plus tier, began voicing deep frustration over the abrupt removal of familiar legacy models like GPT‑4o, along with sudden usage limits and a colder, less engaging tone.
A Wave of Backlash
Many long-time users described GPT‑5 as emotionally distant – “flat,” “uncreative,” even likening it to an “overworked secretary” rather than the warm and personable GPT‑4o they had grown accustomed to. A Reddit discussion, for instance, collected thousands of comments, with one user lamenting, “I truly miss GPT‑4o. It was kind, warm, and always emotionally supportive… GPT‑5 feels robotic and cold,” and threatening to cancel their subscription.
Adding fuel to the fire, the new GPT‑5 system initially capped Plus users at a mere 200 messages per week for the “Thinking” mode—a restriction that many found overly limiting.
OpenAI Responds Swiftly
Despite some mounting criticism, Sam Altman, CEO of OpenAI, took swift and visible action. During a Reddit AMA and on X (formerly Twitter), Altman acknowledged that the team had “underestimated how much some of the things that people like in GPT‑4o matter to them, even if GPT‑5 performs better in most ways”.
He announced a series of key changes to address the uproar:
- Legacy Models Return – GPT‑4o has been restored in the model picker for all Plus users. A new “Show additional models” toggle allows access to variants like GPT‑4.1, o3, and GPT‑5 Thinking mini. GPT‑4.5 remains reserved for Pro users.
- Customizable Modes Introduced – GPT‑5 now offers three distinct “thinking” modes:
- Auto: The model self-selects the ideal approach.
- Fast: Prioritizes speed over depth.
- Thinking (plus mini): Delivers more thoughtful responses, with an expanded context limit of 196 k tokens—ideal for handling long conversations or large documents.
- Dramatic Increase in Usage Limits – The earlier cap of 200 weekly messages in Thinking mode has been raised to 3,000 per week for Plus users. Additionally, base GPT‑5 usage limits have doubled to 160 messages every three hours.
- Personality Tweaks Underway – OpenAI is working to make GPT‑5’s tone “feel warmer” – enough to regain the approachability of GPT‑4o without reverting to a sycophantic style.
- Policy on Model Retirement – Going forward, OpenAI commits to notifying users in advance before retiring older models, acknowledging that emotional attachment matters in AI experiences.
Why These Changes Matter
These rapid changes addressed the most pressing grievance: the loss of familiarity and control. Bringing back GPT‑4o gave users immediate relief. The introduction of multiple thinking modes and vastly expanded message limits empowered users with customizability – no longer forced to trade performance for personality or vice versa.
Moreover, the promise of a warmer GPT‑5 hinted at a more balanced experience—combining technical prowess with engaging tone. And committing to better communication regarding model deprecations shows OpenAI’s awareness of the deeper emotional connections users form with AI.
As a result, the backlash began to subside. Altman’s visible presence and blunt acknowledgement of missteps proved effective in restoring trust and loyalty.
Key Takeaways
| Issue | OpenAI’s Response |
| Loss of model choice | Reintroduced GPT-4o and legacy models for Plus users |
| One-size-fits-all system | Added selectable modes (Auto, Fast, Thinking) for greater flexibility |
| Strict usage limits | Increased message caps to 3,000/wk and 160 messages per 3 hours |
| Cold personality | Working on making GPT-5 feel warmer without overdoing it |
| Abrupt removals | Will now give advance notice before withdrawing models |
In summary, while the GPT-5 launch initially triggered widespread dissatisfaction among ChatGPT Plus users, OpenAI’s rapid and transparent response delivered meaningful solutions – restoring choice, easing limits, and promising a better personality. By tackling users’ top frustrations head-on, OpenAI may well have preserved user engagement and reinforced the importance of emotional resonance when designing AI tools.
