OpenAI has swiftly responded to user criticism of its newly launched GPT-5 model in ChatGPT, implementing significant changes to address concerns about personality, performance clarity, and restrictive usage limits. The updates prioritize user control and accessibility—particularly for ChatGPT Plus subscribers paying $20/month.
Legacy models return following widespread requests. As of this week, GPT-4o—OpenAI’s previous flagship model—is now readily available again for all Plus, Team, and Pro subscribers directly in ChatGPT’s model picker. Users can also enable a new “Show additional models” toggle in settings to access older options like o3, GPT-4.1, and the quicker “Thinking-mini” variant. This reverses OpenAI’s initial approach of limiting choices to prioritize GPT-5. “People have workflows and preferences tied to specific models,” noted tech journalist Amanda Caswell. “Forcing a switch eroded trust.”
Message limits surge after users protested GPT-5’s original 200 weekly messages cap for its deeper-reasoning “Thinking” mode. CEO Sam Altman confirmed a 15-fold increase to 3,000 messages per week for Plus users, with additional capacity via GPT-5 Thinking-mini once exceeded. Pro and Enterprise tiers retain unlimited access. Altman acknowledged ongoing capacity adjustments, stating, “We may have to update rate limits over time depending on usage” as demand climbs.
Granular control replaces guesswork. Users now select between three dedicated modes:
-
“Auto” (default; ChatGPT routes queries automatically)
-
“Fast” (instant responses using lighter models)
-
“Thinking” (deliberate reasoning for complex tasks)
This aims to resolve confusion about when GPT-5 switches underlying models—a pain point highlighted in early feedback. OpenAI is also testing interface indicators showing which model is active during replies.
Personality refinements are imminent. While GPT-5’s current tone was criticized as overly detached, OpenAI promises an upcoming update will make it feel “warmer” without mimicking GPT-4o’s divisively expressive style. Long-term, Altman plans deeper user customization. “Tone is deeply personal,” observed AI researcher Simon Willison. “One-size-fits-all rarely works as models grow capable”.
Compute allocation shifts toward GPT-5 infrastructure after user adoption of reasoning features jumped from 7% to 24% daily among Plus subscribers post-launch. While Pro-tier exclusives like GPT-5 Pro remain, Altman stated expanding capacity for high-demand features is now critical.
OpenAI’s rapid revisions highlight the tension between innovation and user experience in the competitive AI landscape. By restoring choice, expanding access, and refining interaction, the company aims to balance GPT-5’s technical leaps with practical usability. As one developer posted: “It’s about listening, not just launching.”.
Subscribe to my whatsapp channel
Comments are closed.