Artificial Intelligence Update: User Control Enhancements Now Allow Selection of Responses, Making GPT-5 More Responsive to User Preferences
In a bid to improve user satisfaction and address initial criticism, OpenAI has rolled out new features for its flagship language model, GPT-5. The updates, which include the introduction of three response modes - Auto, Rapid, and Thought - aim to provide a more personalized and versatile AI assistant.
The Thought mode, designed for complex tasks like engineering, research, or technical writing, has shown a significant reduction in hallucinations. According to internal data, GPT-5 reduces the hallucination rate sixfold compared to previous models. Responses are about 45% less likely to contain factual errors than GPT-4o, and around 80% less likely when using the "thinking" mode compared to earlier versions.
The Auto mode, on the other hand, automatically routes the query to the most suitable model - fast or deep reasoning - so users do not need to manually select the model type. This improves ease of use and ensures better-tailored responses. The Rapid mode prioritizes speed, sacrificing deep analysis, and works well in casual conversations, small queries, or interactions where immediacy matters more than precision.
OpenAI has also addressed concerns about model reliability, reducing hallucinations, and enhancing reasoning capabilities. GPT-5 is more reliable and helpful in complex, open-ended reasoning tasks, as demonstrated on public factuality benchmarks such as LongFact and FactScore.
Moreover, GPT-5 performs better with 50-80% fewer output tokens across tasks like visual reasoning, agentic coding, and scientific problem-solving. The model also boasts more robust agentic capabilities, improving tools integration such as GitHub Copilot and Microsoft Copilot.
The Thought mode, available for Plus, Team, and Pro users, can process up to 196,000 tokens of context. Sam Altman, CEO of OpenAI, has promised to work on a warmer personality for GPT-5, acknowledging the model's previous cold and robotic demeanor.
The decision to include a switch to show previous versions like GPT-4.1 and o3 is a recognition of the emotional, functional, and ethical value of preserving multiple identities within an AI system. GPT-5 Thought is the favorite for high-risk decisions.
Following the launch, GPT-5 faced some initial issues with model switching and performance inconsistencies. However, six days later, OpenAI partially reversed its architecture, backing away from its total simplification plan with GPT-5. The new system failed at launch, causing massive user frustration and immediate dissatisfaction.
OpenAI has admitted that GPT-5 seemed dumber than expected following its launch. The introduction of the mode selector is an emotional and strategic response to user criticism. The new modes aim to restore user personalization and are a shift in OpenAI's simplification strategy.
Personalization is an essential part of the relationship between humans and AI. The future of AI may lie in systems that allow their behavior to be modulated according to the human context. These updates to GPT-5 are a step towards achieving that goal.
Read also:
- Musk announces intention to sue Apple for overlooking X and Grok in the top app listings
- Innovative Company ILiAD Technologies Introduces ILiAD+: Boosting Direct Lithium Extraction Technology's Efficiency Substantially
- Nuclear Ambitions at a U.S. Airport Spark Controversy, With Opposition Swelling
- Haval H6 Hybrid Analysis: Delving into Engine Performance and Fuel Efficiency
