In my previous article, I examined the anticipated capabilities and potential implications of OpenAI’s GPT-5 model upon its release. Now, a few days into the rollout, it’s clear that the deployment has encountered several challenges. Drawing from recent updates shared by OpenAI CEO Sam Altman on X, this follow-up assesses the current state of affairs, user feedback, and the company’s planned adjustments. I’ll focus on the facts as presented, without speculation.
Key Updates from the Rollout
On August 8, 2025, Altman posted two threads detailing the progress and issues with GPT-5’s deployment. The first update addressed immediate operational fixes and user experience improvements:
- Rate Limits and Model Access: OpenAI plans to double GPT-5 rate limits for ChatGPT Plus subscribers as the rollout completes. Plus users will also have the option to stick with GPT-4o, with the company monitoring usage to determine how long to support older models.
- Performance Adjustments: Altman noted that GPT-5 appeared less capable on the previous day due to a malfunction in the autoswitcher, which determines model selection. Fixes have been implemented to improve model routing, and transparency about which model is in use will be enhanced.
- UI Enhancements: Changes are incoming to make it simpler for users to manually initiate “thinking” modes, potentially improving interaction quality.
- Scale Challenges: The rollout to all users is progressing slower than anticipated due to the model’s scale. API traffic has roughly doubled in the past 24 hours, contributing to delays.
Altman acknowledged that while some “bumpiness” was expected with such a large-scale update, the issues exceeded initial projections.
Later that day, a second update provided further context and forward-looking plans:
- User Preferences for GPT-4o: OpenAI underestimated the attachment users have to certain aspects of GPT-4o, despite GPT-5’s superior performance in many areas. User opinions on the models’ relative strengths vary significantly.
- Customization Needs: Long-term, the company is emphasizing user-specific customization, including research into steerability and model personalities (e.g., preferences for emojis, warmth, or strict logic). This aims to accommodate diverse user needs beyond a one-size-fits-all approach.
- Immediate Priorities: The focus is on completing the rollout (now at 100% for Pro users and nearing completion for others) and making GPT-5 “warmer” in tone. Fuller customization features will take more time.
- Capacity Constraints: A significant capacity shortfall is anticipated for the coming week. Decisions on handling this will be made transparently, though trade-offs may not satisfy everyone.
These updates reflect a responsive approach, with Altman expressing gratitude for user patience and committing to rapid improvements.
User Reactions and Feedback
The responses to Altman’s posts reveal a mix of appreciation, criticism, and specific requests. Common themes include:
- Emotional Attachment to Older Models: Several users emphasized that GPT-4o feels more “warm” or personable, likening the shift to GPT-5 as losing a familiar collaborator. One reply described it as “breaking a bond,” while others highlighted impacts on emotional well-being and called for retaining multiple models indefinitely.
- Performance Concerns: Complaints ranged from GPT-5 feeling “colder” or less engaging to requests for features like multimodal input/output, larger context windows, and access to intermediate checkpoints (e.g., “zenith”).
- Criticism of Rollout Strategy: Some accused the company of overhyping the model or mishandling the transition, such as deprecating previous versions without adequate testing of user migration. Others questioned the autoswitcher as a potential cost-saving measure rather than a user benefit.
- Positive Notes: Amid the critiques, there were acknowledgments of OpenAI’s efforts, such as praise for listening to feedback by reinstating GPT-4o access.
Overall, the feedback underscores a divide: while GPT-5’s raw capabilities are often praised, its rollout has highlighted the importance of qualitative factors like tone, familiarity, and customization in user satisfaction.
Implications and Outlook
The GPT-5 rollout serves as a case study in the complexities of deploying advanced AI at scale. OpenAI’s quick iterations demonstrate adaptability, but the experience also reveals gaps in anticipating user sentiment toward model personalities. Moving forward, the emphasis on customization could address these issues, potentially setting a precedent for more flexible AI ecosystems.
As the situation evolves, I’ll continue monitoring developments. For now, the aftermath of GPT-5’s launch is one of adjustment and refinement, balancing technological advancement with user-centric design.