OpenAI’s new GPT-5 model claims to be a PhD-level expert on anything

Imagine being able to ask a single system to code an application, analyse financial data, explain a complex medical concept, or draft a detailed report in seconds. That is the promise of GPT-5, the latest artificial intelligence model from OpenAI, now available to all 700 million ChatGPT users.

OpenAI is presenting GPT-5 as its most capable model for enterprise use. It excels in software development, finance, health-related queries, and advanced writing.

"This is the first time one of our mainline models feels like you can ask a legitimate, PhD-level expert anything," said CEO Sam Altman. In live demonstrations, GPT-5 generated complete, functioning software from simple text prompts, a process known as "vibe coding".

The release comes at a decisive moment for the AI industry. Technology giants such as Alphabet, Meta, Amazon, and Microsoft have poured billions into AI data centres, expecting major returns. OpenAI itself is exploring a $500 billion valuation, with top researchers receiving signing bonuses as high as $100 million.

However, GPT-5's leap over GPT-4 is debated. Early reviewers praised its coding skills and problem-solving abilities but suggested the improvement was smaller than previous upgrades. Importantly, GPT-5 still cannot learn on its own, keeping it far from human-level intelligence.

One of GPT-5's key innovations is "test-time compute", which allows the model to spend extra processing time on complex problems, improving accuracy in areas such as mathematics and advanced reasoning. This is the first time the public will have access to this feature.