TLDR: OpenAI has rolled out an update for its GPT-5 model, aiming to make its default personality warmer and more approachable. This move comes in response to significant user feedback that criticized the model’s initial formal and robotic tone, with many users expressing a preference for the earlier GPT-4o’s conversational style. The update introduces subtle conversational nuances and new operational modes to enhance user experience.
OpenAI has announced a significant update to its flagship large language model, GPT-5, just weeks after its initial launch. The primary goal of this update is to imbue the AI with a ‘warmer and friendlier’ default personality, a direct response to widespread user feedback. Users had previously criticized GPT-5’s initial release for its overly formal, blunt, and ‘very to the point’ responses, with many expressing a distinct preference for the more approachable tone of its predecessor, GPT-4o.
Sam Altman, CEO of OpenAI, acknowledged the challenges faced during the rollout, stating on X (formerly Twitter) that the launch was ‘a little more bumpy than we’d hoped for.’ He further elaborated on the user sentiment, noting that ‘user attachment to specific AI personalities had been stronger than the company anticipated.’ Altman emphasized that the long-term solution involves allowing users greater customization over ChatGPT’s style, a feature the company is actively developing. He also addressed a temporary issue, stating, ‘GPT-5 will seem smarter starting today. Yesterday, the autoswitcher broke and was out of commission for a chunk of the day, and the result was GPT-5 seemed way dumber.’
Nick Turley, OpenAI’s Vice President, echoed these sentiments, confirming that the initial GPT-5 was indeed ‘just very to the point’ and that the latest update aims to make it ‘more approachable now.’ The changes implemented are described as subtle, incorporating conversational touches such as ‘Good question’ or ‘Great start’ into the AI’s responses. OpenAI assures users that internal tests have shown no increase in sycophancy compared to the previous GPT-5 personality, ensuring the model remains concise while becoming more human-like.
Beyond personality tweaks, the update also introduces new operational modes for GPT-5: ‘Auto,’ ‘Fast,’ and ‘Thinking.’ The ‘Fast’ mode is designed for quicker answers, while ‘Thinking’ mode allows the model more time to generate deeper, more comprehensive responses. The ‘Auto’ mode intelligently switches between Fast and Thinking based on the query. These modes are accessible to users within the model picker in ChatGPT, offering greater control over the AI’s behavior and response style. OpenAI is also working on user interface (UI) improvements to make these modes more intuitive to select and to clearly indicate which AI model is responding.
Also Read:
- OpenAI’s O3 Model Surpasses Newer GPT-5 in Complex Multi-Application Office Workflows
- OpenAI’s Sam Altman Discusses AI’s Future Beyond GPT-5, Including Brain-Computer Interfaces and Societal Integration
The GPT-5 model, which became freely available in ChatGPT this month after its August 7 launch, had faced some disappointment from users who felt the advancements were incremental despite significant hype and previous delays due to safety testing and compute limitations. This update underscores OpenAI’s commitment to refining user experience and adapting its AI models based on real-world interaction. Data from platforms like IndexBox indicate a rising adoption of AI language models across industries, with enterprises increasingly prioritizing conversational ease and adaptability in their AI interactions, making this update particularly relevant to the broader AI landscape. The updated GPT-5 is now also available for ChatGPT Enterprise and Edu subscribers.
