ChatGPT revision: Altman details three modifications to GPT-5, following user criticism
In a move to cater to the preferences of its users, OpenAI has made several significant changes to its flagship product, ChatGPT, following the launch of GPT-5 earlier this month.
CEO Sam Altman acknowledged the unexpected backlash from long-time ChatGPT users, who expressed their attachment to the GPT-4o model. In response, OpenAI has brought back GPT-4o for Plus and Pro subscribers, recognising that AI upgrades can't ignore user preference.
The first change in the updated ChatGPT experience is the introduction of three selectable response modes: Auto, Fast, and Thinking. The Auto mode balances speed and depth depending on the query, while the Thinking mode allows GPT-5 to take extra time to process, aiming for deeper reasoning, more nuance, and more carefully structured answers. The Fast mode prioritises speed above all else.
OpenAI has also addressed concerns about reduced user control and the forced use of GPT-5 for all users. The toggle now allows access to other non-default models, including o3, 4.1, GPT-5 Thinking mini, and GPT-4o for Plus and Pro subscribers.
To improve the user experience, OpenAI has refined GPT-5's style to be less effusive, use fewer unnecessary emojis, and produce responses more like chatting with a knowledgeable friend than a robotic AI. The company has also reduced excessive agreeableness and flattery seen in earlier versions like GPT-4o.
In addition, ChatGPT has been enhanced to better decide when and how to use tools (web search, Python, image generation) to handle complex multi-step queries more effectively in under a minute. This represents a move towards more agentic AI that can execute multi-faceted tasks independently.
OpenAI has also introduced a low-cost subscription, ChatGPT Go, initially in India, to improve accessibility. GPT-4o’s memory and problem-solving abilities for STEM tasks have been fine-tuned, making it more proactive in guiding conversations.
For Plus subscribers, the rate limits have been doubled, resulting in higher capacity. Plus subscribers get around 3,000 messages per week in the Thinking mode, after which they can use a lighter "Thinking mini" model for extended reasoning without the same computational cost.
Lastly, OpenAI has added new GPT-5 API features for developers, including control over verbosity and reasoning effort, multiple model sizes for cost/latency trade-offs, and support for custom tools callable by plaintext. These give developers more flexibility and control over interactions.
OpenAI is actively addressing user concerns from the GPT-5 rollout, mainly by adjusting the model router to allow better access to preferred models, reducing overly sycophantic behavior, and enhancing ChatGPT’s reasoning & multi-tool integration to improve usefulness and user satisfaction.
Read also:
- Musk announces intention to sue Apple for overlooking X and Grok in the top app listings
- Nuclear Ambitions at a U.S. Airport Spark Controversy, With Opposition Swelling
- Haval H6 Hybrid Analysis: Delving into Engine Performance and Fuel Efficiency
- Veolia advocates for sustainability by financing eco-friendly environmental projects
