LONDON (The Thursday Times) — OpenAI has launched GPT-5, the latest iteration of its flagship artificial intelligence model, hailing it as a “significant step” towards artificial general intelligence (AGI) while acknowledging that the technology is still far from reaching the autonomy and adaptability required to match human capability.
The upgrade, which is now being rolled out to all of ChatGPT’s 800 million users, introduces major improvements in areas such as coding, creative writing, factual accuracy and user interaction. It also includes more nuanced safety responses and new capabilities such as integration with Gmail and Google Calendar via the ChatGPT agent feature.
Sam Altman, OpenAI’s chief executive, described GPT-5 as “generally intelligent” and a “huge improvement” on its predecessors. However, he conceded that the system does not yet meet the company’s own definition of AGI — a highly autonomous tool that can outperform humans at most economically valuable tasks.
“I think the way that most of us define AGI, we’re still missing something quite important, many things quite important,” Altman said, adding that one of the biggest shortcomings was the model’s inability to learn continuously after deployment. “That feels like it should be part of AGI.”
Despite these limitations, GPT-5 has shown progress on several fronts. OpenAI says the model makes fewer factual errors, performs better at writing functional code, and shows greater ability in creative tasks such as storytelling and poetry. It is also less sycophantic — a reference to the chatbot’s previous tendency to overly agree with users, which raised concerns about its credibility and emotional safety.
Nick Turley, head of ChatGPT at OpenAI, confirmed that GPT-5 includes specific guardrails to curb sycophancy, following feedback that the overly agreeable tone of previous versions could confuse or distress users.
The system has also been fine-tuned to handle sensitive topics more responsibly. Rather than bluntly refusing to answer prompts that breach safety guidelines, GPT-5 now attempts to respond constructively within those limits or provides clearer explanations for why it cannot help.
Health-related queries will see particular improvement. OpenAI says GPT-5 will be more proactive in flagging potentially serious issues — such as signs of mental illness — but emphasised that the tool is not a replacement for professional medical advice, amid ongoing concerns about the risks AI tools may pose to vulnerable individuals.
The release comes amid intensifying competition among major technology firms to develop AGI. Google’s AI division this week showcased an experimental “world model” as part of its own efforts, while Meta CEO Mark Zuckerberg recently stated that the advent of superintelligence — a more advanced theoretical form of AI than AGI — was “now in sight”.
The race to achieve AGI has sparked intense debate about the future of human employment. Dario Amodei, chief executive of rival AI developer Anthropic, warned earlier this year that AI could replace half of all entry-level white-collar jobs within five years.
OpenAI’s ambitions have made it one of the most highly valued private companies in the world. It was reported this week that the San Francisco-based firm is in early talks to sell shares held by employees and ex-staff at a valuation of $500 billion — surpassing the estimated worth of Elon Musk’s SpaceX.
Although OpenAI recently released two smaller open-source models and continues to offer a free version of ChatGPT, access to GPT-5 will be limited unless users upgrade to a paid plan. The $200-per-month ChatGPT Pro package will grant full access to the new model, while free-tier users will face usage caps.
With the GPT-5 rollout, OpenAI has again raised the stakes in the global race for AGI — but even as the tools grow smarter and more powerful, its own leadership admits the finish line is not yet in sight.