GPT-5 is here! But is it any good?

GPT-5 was released on the 7th August 2025, and has been making headlines in the tech world.

In this episode of the Inspiring Tech Leaders podcast, I explore the following: 

💡 GPT-5 now "thinks before it speaks" with internal chain-of-thought processing

💡 Massive reliability boost with a 26% reduction in hallucinations vs GPT-4o

💡 Strongest coding model yet, creating complete apps from simple descriptions

💡 Available to free users for the first time, breaking down barriers to advanced AI

The competitive landscape is heating up with bold claims from all players, but GPT-5's step-by-step reasoning and improved accuracy might just be what we've all been waiting for.

The big question - Are we witnessing the moment AI transitions from sophisticated tool to intellectual collaborator?

What's your take on GPT-5?  Have you tried it yet?

Available on: Apple Podcasts | Spotify | YouTube | All major podcast platforms

Start building your thought leadership portfolio today with INSPO.  Wherever you are in your professional journey, whether you're just starting out or well established, you have knowledge, experience, and perspectives worth sharing. Showcase your thinking, connect through ideas, and make your voice part of something bigger at INSPO - https://www.inspo.expert/

Chapters

• 0:00 Introduction
• 2:28 Revolutionary Improvements
• 4:48 Healthcare and Specialised Domains
• 6:35 The Three-Tier Approach: GPT-5, Mini, and Nano
• 8:28 Democratising Access
• 9:44 AI Safety and Ethical Considerations
• 11:02 The Implications of PhD-Level AI
• 12:11 The Sceptical Perspective
• 12:46 The Regulatory Challenge
• 14:55 Wrap Up

Transcript

Welcome to the Inspiring Tech Leaders podcast, with me Dave Roberts.  This is the podcast that talks with tech leaders from across the industry, exploring their insights, sharing their experiences, and offering valuable advice to technology professionals.  The podcast also explores technology innovations and the evolving tech landscape, providing listeners with actionable guidance and inspiration.

This week I’m taking a closer look at the recent introduction of OpenAI’s GPT-5.  I will be evaluating what’s new in the latest model and the response it’s received since its introduction.

Released on 7th August, 2025, OpenAI unveiled what CEO Sam Altman is calling a major upgrade, and is what is touted as the smartest, fastest and most advanced AI model to-date with 700 million weekly active users.

If you've been following the AI space, you know that each new generation of GPT models has brought remarkable improvements.  But GPT-5 isn't just an incremental update, it’s being positioned as the first AI that truly feels like talking to a PhD-level expert in any field.

So, let’s start with the headline that's capturing everyone's attention.  OpenAI claims that GPT-5 represents the first time an AI truly feels like conversing with a PhD-level expert.  Sam Altman has said the GPT-5 is a significant step forward, describing how GPT-3 felt like you were talking to a high school student, impressive for its time, but limited in depth and reasoning.  GPT-4, he said, felt like conversing with a college student, more sophisticated, better at complex tasks, but still lacking the deep expertise and nuanced understanding we expect from true experts.

GPT-5, according to Altman, crosses that crucial threshold.  GPT-5 is the first time that it really feels like talking to an expert in any topic, like a PhD-level expert, he explained.  This isn't just about having more knowledge, it’s about the quality of reasoning, the ability to think through complex problems, and the capacity to provide insights that demonstrate genuine understanding rather than sophisticated pattern matching.

The implications of this leap are profound.  We're potentially looking at an AI that can engage in the kind of deep, nuanced discussions that were previously the exclusive domain of human experts.  Whether you're a researcher grappling with complex scientific problems, a developer working on intricate coding challenges, or a professional seeking expert-level analysis, GPT-5 promises to be a collaborator rather than just a tool.

So, what exactly makes GPT-5 so much more capable than its predecessors? The improvements span multiple dimensions, each contributing to what OpenAI describes as a model that's smarter, faster, and more useful.

One of the most significant improvements is in the area of reasoning and accuracy.  Essentially it thinks before it speaks.  Unlike previous models that would generate responses immediately, GPT-5 carries out an internal chain of thought, working through problems step by step before providing its final answer.

This approach has yielded remarkable results in reducing one of AI's most persistent problems, hallucinations.  When AI models "hallucinate," they confidently present false information as fact, a critical issue that has limited the reliability of AI systems.  OpenAI's testing shows that GPT-5 has a 26 percent reduction in hallucination rate compared to GPT-4o, and the thinking version of GPT-5 shows an even more impressive 65 percent reduction compared to the o3 model.

But perhaps more importantly, GPT-5 has been trained to handle uncertainty more gracefully.  Instead of making up answers when it doesn't know something, it's been designed to "fail gracefully”, recognising when a task can't be completed, avoiding speculation, and explaining its limitations more clearly.  This represents a fundamental shift toward more a honest and reliable AI interaction.

For developers and programmers, GPT-5 represents what might be the most significant advancement yet in AI-assisted coding.  OpenAI is positioning it as their strongest coding model to date, with particular improvements in complex front-end generation.

The model excels in what OpenAI calls "vibe coding”, the ability to generate complete software applications from simple, natural language descriptions.  During the launch demonstration, researchers asked GPT-5 to create a web app to help English speakers learn French, specifying that it should have an engaging theme, include activities like flashcards and quizzes, and track daily progress.  Within seconds, GPT-5 generated not one, but two different complete applications that met all the specified requirements.

This is not just an incremental improvement, it represents a fundamental shift forward in how AI can assist with software development.  GPT-5 isn't just writing code, it's understanding complex requirements, making architectural decisions, and creating complete, functional applications.

OpenAI is also highlighting GPT-5 as their best model yet for health-related questions.  In specialised healthcare benchmarks, GPT-5 outperformed previous models by substantial margins.

What makes these healthcare improvements particularly noteworthy is that the scores are validated by multiple physicians, ensuring that the AI's medical reasoning aligns with professional medical standards.  This suggests that GPT-5 could become a valuable tool for medical professionals, researchers, and even patients seeking reliable health information.

To truly appreciate the significance of GPT-5, it's important to understand the underlying technology that makes these advances possible.  At its core, GPT-5, like its predecessors, is built on what's called the transformer architecture, a revolutionary approach to processing language that has fundamentally changed the field of artificial intelligence.

The transformer architecture, introduced in 2017, represents a paradigm shift in how machines process language.  Unlike earlier approaches that processed text sequentially, word by word, transformers use something called self-attention mechanisms that allow the model to consider the relationship between every word and every other word in a text simultaneously.

Think of it this way, when you read a sentence, you don't just process each word in isolation.  You understand how each word relates to all the others, how the meaning of one word might change based on words that come much later in the sentence, and how the overall context shapes the interpretation of individual elements.  The transformer architecture gives AI models a similar capability.  This self-attention mechanism is what allows GPT models to maintain context over long conversations, understand complex relationships between ideas, and generate coherent responses that take into account the full scope of what's been discussed.

OpenAI is releasing GPT-5 in three different versions through their API, these being GPT-5, GPT-5-mini, and GPT-5-nano.  This tiered approach reflects the reality that different applications have different requirements depending cost, speed, and capability.

GPT-5 represents the full-capability mode, offering the complete range of reasoning and advanced features we've discussed.  GPT-5-mini provides a more cost-effective option that maintains much of the capability while being faster and less expensive to run.  GPT-5-nano offers the most streamlined version for applications where speed and cost are the primary concerns.

This approach allows developers and organisations to choose the right balance of capability and efficiency for their specific needs, making advanced AI more accessible across a broader range of applications.

GPT-5's launch comes at a time of intense competition in the AI space, with major players making increasingly bold claims about their models' capabilities.  Just last month, Elon Musk claimed that his AI chatbot Grok was "better than PhD level in everything" and called it the world's "smartest AI”.  This kind of competitive rhetoric highlights how rapidly the field is advancing and how high the stakes have become.

The competition isn't just about bragging rights, it’s about establishing dominance in what many see as the most important technological development of our time.  Anthropic, with their Claude models, has been targeting similar markets, particularly in coding assistance.  The rivalry has even led to some tension, with Anthropic recently revoking OpenAI's access to their API, claiming violations of terms of service.

This competitive environment is ultimately beneficial for users and developers.  The pressure to outperform rivals is driving rapid innovation, pushing each company to develop more capable, more reliable, and more useful AI systems. GPT-5's impressive capabilities are, in part, a response to this competitive pressure.

One of the most significant aspects of the GPT-5 launch is OpenAI's decision to make it available to all users, including those on the free tier.  This represents a notable shift in strategy, previous cutting-edge models were typically reserved for paying subscribers, at least initially.

The availability structure is designed to accommodate different user needs and usage patterns.  Free users get access to GPT-5 with usage caps, after which they can access GPT-5-mini.  Plus subscribers enjoy higher usage limits, while Pro subscribers get unlimited access to GPT-5 and access to the even more advanced GPT-5 Pro version.

This democratisation of access is particularly significant for the reasoning capabilities.  GPT-5 marks the first time that free users have access to a reasoning model, the kind of AI that can think through problems step by step. This could have profound implications for education, research, and innovation, as advanced AI capabilities become available to a much broader audience.

For enterprise users, GPT-5 is being integrated into Microsoft's ecosystem, including Microsoft 365 Copilot and Azure AI Foundry.  This integration means that millions of business users will soon have access to GPT-5's capabilities directly within their existing workflows and applications.

With great capability comes great responsibility, and OpenAI has invested heavily in safety research for GPT-5.  The company conducted over 5,000 hours of safety testing during development, focusing on reducing harmful outputs and improving the model's ability to handle sensitive topics appropriately.

One notable improvement is in how GPT-5 handles potentially risky questions.  Instead of simply refusing to answer, GPT-5 uses "safe completions”, providing high-level responses within safety constraints that can't be used to cause harm.  This approach maintains the model's helpfulness while reducing potential risks.

OpenAI has also worked to reduce what they call "deception" in GPT-5, the model's propensity to mislead, cheat, or find inappropriate shortcuts to solve problems.  While the company acknowledges that their mitigations aren't perfect and more research is needed, they've made significant progress in training the model to be more honest and transparent about its limitations.

The company is also promoting healthier relationships between users and AI.  Recent updates include changes to how ChatGPT handles personal advice, moving away from giving definitive answers to sensitive questions and instead helping users think through problems by asking questions and weighing pros and cons.

As we consider the broader implications of GPT-5, we're looking at a technology that could fundamentally change how we work, learn, and solve problems.  The availability of PhD-level expertise on demand has the potential to accelerate research, democratise access to expert knowledge, and enable new forms of human-AI collaboration.

For researchers and academics, GPT-5 could serve as a tireless research assistant, capable of analysing complex data, identifying patterns, and suggesting new avenues for investigation.  For students, it could provide personalised tutoring at an expert level, adapting to individual learning styles and needs.  For professionals across industries, it could offer specialised expertise that was previously accessible only through expensive consultants or years of specialised training.

However, this advancement also raises important questions about the future of expertise and professional knowledge work.  If AI can provide PhD-level insights across multiple domains, how does this change the value proposition of human experts?  The answer likely lies not in replacement, but in augmentation, AI enabling human experts to work at higher levels of abstraction and tackle more complex challenges.

However, not everyone is convinced that GPT-5 represents the revolutionary leap that OpenAI claims. Some have described GPT-5 as an evolution rather than a revolution and think that there is a lot marketing hype to maintain investor enthusiasm.

There are others that have been even more critical of GPT-5, saying that it’s been underwhelming based on all the hype surrounding the launch of this latest model.

These sceptical voices provide important balance to the enthusiasm surrounding GPT-5.  They remind us that technological progress often feels more gradual to users than the dramatic leaps suggested by benchmark improvements and corporate announcements.

However, as AI capabilities continue to advance, the challenge of governance and regulation becomes increasingly urgent.

The gap between AI capabilities and our ability to govern them effectively is widening.  GPT-5's advanced reasoning abilities, while beneficial, also raise new questions about accountability, transparency, and control.  How do we ensure that AI systems with PhD-level capabilities are used responsibly?  How do we maintain human agency and oversight when AI can outperform humans in many cognitive tasks?

These questions don't have easy answers, but they're becoming increasingly important as AI capabilities continue to advance.

GPT-5 represents more than just another incremental improvement in AI technology.  It marks a potential inflection point, the moment when AI truly begins to feel like conversing with an expert rather than a sophisticated tool.  The combination of enhanced reasoning capabilities, reduced hallucinations, improved coding abilities, and specialised domain expertise suggests that we're now entering a new era of human-AI interaction.

Whether GPT-5 lives up to its ambitious promises remains to be seen.  The true test will come as millions of users begin incorporating it into their daily workflows, pushing its capabilities to the limits and discovering both its strengths and limitations.

What's clear is that the pace of AI development shows no signs of slowing.  If GPT-5 represents PhD-level expertise, what might GPT-6 or GPT-7 bring?  The trajectory suggests we're moving toward AI systems that don't just assist with human tasks, but actively collaborate as intellectual partners.

For now, GPT-5 offers us a glimpse of that future, a world where expert-level AI assistance is available to anyone with an internet connection, where complex problems can be tackled with AI reasoning capabilities, and where the boundaries between human and artificial intelligence continue to blur in fascinating and sometimes unsettling ways.

As Sam Altman noted, people are limited by ideas, but not really the ability to execute, in many new ways.  GPT-5 might just be the tool that helps us bridge that gap between imagination and implementation.

GPT-5 is rolling out now to users worldwide, and I'll be watching closely to see how this new generation of AI performs in the real world.  Will it live up to the PhD-level promises?  Only time will tell.  

Well, that is all for today.  Thanks for tuning in to Inspiring Tech Leaders. If you enjoyed this episode, don’t forget to subscribe (it actually does help), leave a review, and share it with your network.  You can find more insights, show notes, and resources at www.inspiringtechleaders.com

Head over to the social media channels, you can find Inspiring Tech Leaders on X, Instagram, INSPO and TikTok.  Let me know your thoughts on GPT-5.

Thanks for listening, and until next time, stay curious, stay connected, and keep pushing the boundaries of what is possible in tech.