ChatGPT Upgrade Promises Fewer Hallucinations and Greater Health Query Precision

OpenAI has unveiled the latest iteration of its flagship chatbot, GPT5, marketed as the most advanced model yet in reducing incorrect or fabricated responses. The company claims this upgrade will greatly diminish the prevalence of so-called ‘hallucinations’—instances where the chatbot invents answers—an issue that has recently raised concerns within both the consumer and regulatory landscape.

According to OpenAI, the most capable version of GPT5 is now eight times less likely to make mistakes on health topics than the company’s prior advanced release and outperforms the free version by a factor of fifty. These gains stem from rigorous testing on a set of five thousand simulated health-related queries, aiming to mirror conversations patients might have with medical professionals.

The model is now designed to proactively flag potential health concerns based on the user’s conversation, foreshadowing the broader role artificial intelligence is set to play in medicine. While OpenAI stresses its chatbot should not replace qualified clinicians, improved proficiency in managing health queries underscores the growing confidence in AI as a tool for both patients and providers. With healthcare costs rising and patient loads increasing, such technology could become an essential force multiplier for strained public health systems like the NHS.

GPT5’s upgrade is not restricted to medical competence. The model now demonstrates the capability to write complex software from minimal instruction—a feat termed ‘vibe coding’ by its developers. For instance, an engineer was able to design a French language learning mini-game, reminiscent of the iconic mobile game Snake, in under five minutes, using only a few lines of instruction in English.

Sam Altman, chief executive of OpenAI, describes the latest system as a “significant step along the path to AGI”—artificial general intelligence—a point where computers are expected to match or surpass human intellect. The release has stoked debate in financial circles over the potential for widespread job disruption and the acceleration of technological progress, as AI models cross new frontiers at an accelerating pace.

Roughly seven hundred million people now use OpenAI’s services weekly, indicating vast popular uptake and an ever-greater influence in both commercial and consumer sectors. OpenAI also notes the new model has been engineered to be less sycophantic, answering user queries with greater directness and less unnecessary flattery—a nuance welcomed by professionals seeking sober and accurate analysis.

As investment in AI reaches record levels and governments like the UK’s intensify their push for technological modernisation within key public sectors, the race for stable, reliable AI is firmly underway. OpenAI’s latest step demonstrates how quickly generative models are closing the gap with human expertise, bringing both opportunity and challenge for the worlds of finance, medicine and public policy.