ChatGPT-maker OpenAI has unveiled the long-awaited latest version of its artificial intelligence (AI) chatbot, GPT-5, with the claim that it can provide PhD-level expertise.
CEO Sam Altman confidently described the upgrade as akin to “talking to a PhD-level expert in anything, any subject” at Thursday’s launch.
“I believe having something like GPT-5 would have been nearly inconceivable at any other point in human history,” Altman declared.
OpenAI describes GPT-5 as “smarter across the board,” with marked improvement in its ability to code, write, reason, and answer questions spanning fields from mathematics to healthcare. The model now delivers “more useful responses across maths, science, finance, law and more – like having a team of experts on call for whatever you want to know,” according to the official statement.
Independent tests echo this. “GPT-4, maybe it was like talking to a college student. But with GPT-5, now it’s like talking to an expert, a PhD-level expert in anything,” a developer told Sky News.
Benchmarks reveal near-perfect results in competitive mathematics and robust performance in coding and writing tasks—areas traditionally distinguished by advanced academic expertise. On specialized competitions like the Harvard-MIT Math Tournament, GPT-5 reportedly achieved 100% accuracy when paired with computational tools, and even its code generation abilities now rival experienced professionals.
Smarter, Safer, and More Reliable?
OpenAI is touting more than brains – GPT-5 is also being called its safest model ever. Altman emphasised the practical benefits: “GPT-5 not only surpasses previous models in performance metrics and responds more swiftly, but – most critically – proves to be more beneficial for practical inquiries”.
Safety features have evolved, too. The hallucination rate (incidence of fabricated answers) is down dramatically from previous versions—a core concern for AI skeptics. “GPT-5 has been trained to identify when a task cannot be completed, steer clear of speculation, and articulate its limitations more transparently, which minimizes unsupported claims compared to earlier models,” said Pokr, OpenAI’s training lead. This output-centric safety means users get as helpful an answer as possible—within the guardrails of safe, factual AI.
Is “PhD-Level” a Marketing Term or a Measurable Standard?
While the “PhD-level” phrase grabs attention, it’s as much about positioning as precision. Experts and commentators point out that true PhD expertise involves critical thinking, deep research skills, and domain-specific judgment—still a high bar for any AI. OpenAI, however, cites specific improvements:
- Coding: More complex projects handled end-to-end.
- Mathematics: Solves elite-level problems with high accuracy.
- Writing & Reasoning: Offers more nuanced, contextually accurate responses.
Still, as one analyst noted: “The public rollout and the company’s tone immediately prompted both excitement and skepticism from the public, regulators and academic experts who asked whether marketing matched measurable capability”.
What’s Next – and Can You Try It?
GPT-5 is now available to all ChatGPT users—even free users—with paid tiers unlocking extra features and enhanced capabilities. Microsoft is integrating the model across its platforms, allowing enterprise users and developers to leverage new reasoning and coding powers.
For now, the tech world will be watching closely to see if ‘Dr. ChatGPT’ can deliver on its high degree of hype.
