OpenAI has once again redefined the boundaries of artificial intelligence (AI) with the introduction of GPT-5, its most powerful and versatile AI model to date, with CEO Sam Altman saying it can provide “PhD-level” expertise in areas such as writing, coding, and in healthcare.
Combining unprecedented speed, precision, and adaptability, the release of GPT-5 has set a new industry standard by seamlessly handling everything from everyday queries to complex, expert-level tasks.
“GPT-3 sort of felt to me like talking to a high school student… 4 felt like you’re kind of talking to a college student,” Altman said in a briefing ahead of the launch on Thursday. “GPT-5 is the first time that it really feels like talking to an expert in any topic, like a PhD-level expert.”
As it was announced last month, OpenAI promised that GPT-5 will come with a “unified” model. And it indeed came with it, making it the most important feature.
Unlike previous versions, which required users to pick between different AI models tailored to fast responses or deep reasoning, GPT-5 automatically determines the best approach for each task. It achieves this through a “real-time router,” a system that analyzes the complexity of user requests or prompts and seamlessly switches between models designed for either quick answers or detailed problem-solving.
This approach means users no longer need to understand or choose between different AI models, as it is now left to the new model to choose what it deems best for the task at hand. Whether it is answering a simple question or handling complex professional tasks, GPT-5 instantly adapts to deliver the best results.
GPT-5 outstanding performance across key areas
GPT-5 excels notably in several core areas like in coding, in healthcare and science, in mathematics, and in writing. Here is how the model performs in these sectors:
- Coding and Software Development: The new model scored a 74.9% on a complex software engineering benchmark (SWE-bench), significantly outperforming GPT-4 and other earlier models. It also achieved a 88% on Aider Polyglot, a coding benchmark to evaluate how well different large language models (LLMs) can solve coding tasks across multiple programming languages. GPT-5’s coding capabilities now include generating full applications from simple prompts, managing multi-file projects, and accurately debugging complicated code bases. During demonstrations, GPT-5 rapidly built a Rolling ball minigame with a single prompt, making it a game-changer for software developers.
- Healthcare and Science: In health-related queries and scientific applications, GPT-5 has so far shown improved safety and accuracy. While it maintains strong caution regarding medical advice, its tailored updates and protocols make it more reliable for healthcare professionals seeking assistance from AI-powered analytics. On HealthBench Hard, an open-source benchmark specifically built to evaluate large language models’ (LLMs) performance in realistic medical contexts, healthcare on GPT-5 achieved 46.2%.
- Mathematics: GPT-5 has also reached new heights in mathematical problem-solving. It achieved a 94.6% without tools on the American Invitational Mathematics Examination (AIME 2025), a benchmark no previous AI has reached. This was accomplished using integrated Python tools, and it showcases GPT-5’s ability to tackle challenging computations with precise accuracy.
- Writing: According to OpenAI, GPT-5 is their “most capable writing collaborator yet.” It is now capable of shaping rough ideas into clear, polished prose. It also can better handle routine tasks like drafting emails, reports and memos, while also reliably producing structurally complex creative work. An example is generating an unrhymed iambic pentameter (blank verse), something the prior models couldn’t properly do. Overall, GPT-5 is reported to deliver more consistent tone, rhythm and form than former models.
Safer and more reliable AI interactions
With advancements in AI come concerns about safety. OpenAI has addressed these with new features that reduce hallucinations, where AI-powered chatbots produce incorrect or misleading information. Now, according to OpenAI, “GPT‑5’s responses are ~45% less likely to contain a factual error than GPT‑4o, and when thinking, GPT‑5’s responses are ~80% less likely to contain a factual error than OpenAI o3.”
Another breakthrough is the introduction of “more honest responses,” a smarter way for the chatbot to handle sensitive or potentially dangerous prompts. Rather than simply refusing to answer, or even answering and feeding into the dangerous prompts, GPT-5 provides helpful, high-level guidance while staying within strict safety boundaries. It explains why some details cannot be given and suggests safer alternatives when appropriate.
However, the safety concerns still remain. On the one hand, some experts think the development of AI models to effectively cater to the needs of its users might just be a marketing bubble that cannot be sustained for a long time. “These systems, as impressive as they are, haven’t been able to be really profitable. There is a fear that we need to keep up the hype, or else the bubble might burst, and so it might be that it’s mostly marketing,” Prof Carissa Véliz of the Institute for Ethics in AI said.
On the other hand, there are industry experts and analysts calling for more regulation as the development of AI intensifies. “As these models become more capable, the need for comprehensive regulation becomes even more urgent,” warned Gaia Marcus, a Director of the Ada Lovelace Institute — an independent research and deliberative body dedicated to ensuring that data and AI technologies work for the people and society.
Grant Farhall, Chief Product Officer at Getty Images also raises concerns about the protection of creators’ intellectual property and the compensation of their work. “As AI content becomes more convincing, we need to ask ourselves — are we protecting the people and creativity behind what we see every day?” he asked. “Authenticity matters — but it doesn’t come for free.”
Competitive landscape and AI evolution
While GPT-5 leads in performance and cost-effectiveness, the AI field remains highly-competitive. Models like Anthropic’s Claude Opus 4.1 and Google’s Gemini 2.5 Pro offer strong competition, particularly in specialized coding and ecosystem integration. But GPT-5’s balance of safety and flexibility allows it to become a preferred choice among the general users of AI-powered chatbots.
With GPT-5, OpenAI firmly positions itself at the forefront of AI innovation, where it offers a tool that is smarter, faster, and more accessible than ever before. As AI continues to evolve, GPT-5 represents an important milestone on the path to developing increasingly intelligent and practical artificial intelligence.
