Over the past few days I’ve had only GPT-5 and GPT-5 (Thinking), and both have been terrible. Context retention is nearly nonexistent—key project details get lost just a few turns later.
Making matters worse, OpenAI removed the 4o, 4.1 and 4.5 models, all of which handled context far better than GPT-5. Honestly, GPT-5 is the worst ChatGPT experience I’ve had in a long time.
What say you, anyone else having issues with ChatGPT 5?
Follow-Up: OpenAI Caves to Backlash – GPT-4o Returns
Yay! After massive user backlash—especially from experienced developers, power users, and professionals—OpenAI has quietly restored access to the GPT-4o model in the ChatGPT interface for Plus subscribers.
History: Just days ago, GPT-5 replaced all previous models, and users had no way to select GPT-4o, 4.1, or 4.5. The reaction was immediate and widespread:
Threads on Reddit, Hacker News, and OpenAI’s own forums exploded with complaints.
Multiple media outlets picked up the story, including Ars Technica, Hindustan Times, and Futurism.
Feedback focused on GPT-5’s inability to maintain basic context, its degraded reasoning in technical threads, and its "chatty-cathy" engagement style that annoyed users trying to get real work done.
As of today, GPT-4o is back—quietly reintroduced in the model selector. No major announcement. No public apology. Just damage control.
Presently, Plus users must go into their Settings and enable “Legacy Models” to get the 4o option.
I’ve tested it, and yes, GPT-4o performs far better in terms of context retention, situational awareness, and concise technical reasoning. If you're a Plus subscriber doing anything serious with AI—engineering, systems analysis, coding, etc., which requires retaining context from prior chats or project files—switch back to 4o.
This is a clear case of OpenAI misjudging its user base. Fortunately, the outcry worked, at least partially. But this episode raises valid concerns about forced model migrations and lack of transparency from a company with so much control over mainstream AI use.
I consider myself a “Power ChatGPT User” and (after a few days of testing) found ChatGPT 5 to be a serious downgrade—What in the world was OpenAI thinking?
It was in the newspaper yesterday…
OpenAI brought back four models, but it still defaults to version 5 and there’s no way to set a different model as the default.
Upside: the four models are back. Downside: OpenAI showed rare stupidity by removing all models except 5 with zero notice to users.
This is how the people running the world’s most widely used LLM-based AI operate. - We are doomed! (hahaha)
OODA on Altman/OpenAI Team Competence Post-GPT-5 Backlash
Massive user backlash: Tens of thousands of Reddit posts, Hacker News threads, and media articles blasting GPT-5’s regression in context, reliability, and “personality” .
Customer migration: Some business and power users temporarily stopped or reduced usage, citing productivity loss .
Emergency rollbacks: OpenAI restored legacy models (GPT-4o, o3, o4-mini, 4.1) only after widespread outrage, with zero meaningful transparency or pre-warning .
Leadership fumbles: Sam Altman’s social posts and public comments have been defensive, promising “personality tweaks,” “rate limit increases,” and “we’ll notify before changes” only after the incident.
Industry press panned the rollout: Media called it “rushed,” “sloppy,” “out of touch with users,” and compared the product to a “corporate beige zombie” .
Pattern of poor change management:
No staged rollout, feedback loop, or rollback plan—legacy models removed overnight.
Insufficient communication and zero documentation of planned changes for users and devs.
Lack of risk assessment:
AI platforms are sticky; users build habits, workflows, and emotional attachment to models.
Altman’s team failed to anticipate the depth of user reliance or the negative reaction to forced upgrades and loss of model diversity.
Reactive leadership:
All “fixes” came after major user revolt.
Patches (legacy model return, personality tuning) are reactionary, not strategic.
Altman and his executive/product team demonstrated severe incompetence in user relations, change management, and product leadership during this GPT-5 transition.
This was not just a technical glitch—it was a complete failure in understanding user needs, emotional investment, and basic risk mitigation in platform upgrades.
For OpenAI (if they want to regain trust):
Establish proper change management and staged rollout protocols.
Communicate planned changes to all user tiers—early, openly, and with options.
Involve power users and businesses in beta/feedback cycles.
Document a clear rollback strategy for major releases.
Stop gaslighting: Admit missteps, not just promise “tweaks.”
For users:
Assume all future OpenAI changes could be abrupt and user-hostile.
Build redundancy and have migration options if you rely on specific models/workflows.
Summary:
This incident is textbook tech leadership failure. If this were a traditional SaaS company, the board would be demanding heads roll. Altman and team ran OpenAI like a VC-funded beta project, not a production platform with a real user base.
References:
Direct, no-nonsense. This is how you lose user trust—and how you demonstrate you don’t deserve to run a mission-critical platform.
