Through my training and consultancy work, I’m in daily contact with ChatGPT – it’s long since become something like a reliable sparring partner for me. When GPT-5 was released in Germany last Thursday night, I wasted no time and integrated the new model straight into my workflow. My first hours with it were a mix of anticipation and scepticism: could an already strong version like GPT-4o really make a noticeable leap forward? After several intense days of use, my answer is yes – and in such a way that I already wouldn’t want to be without it.
The most obvious improvement is precision. In my projects, I design training programmes, analyse trends, and develop AI strategies for businesses. GPT-5 now hits the mark far more consistently, structures its answers more clearly, and produces fewer factual missteps. Especially when evaluating complex market developments or drafting strategic options, it’s a case of fewer detours, more bull’s-eyes. Then there’s the speed: where GPT-4o sometimes took its time, results now arrive in about half the time, crisply formulated and free from unnecessary waffle.
Another strong point in my day-to-day work is the noticeably reduced tendency towards “hallucinations”. Incorrect information hasn’t vanished entirely, but the rate has dropped significantly – an important factor when reliability matters. Particularly valuable is the expanded “reasoning” capability – the ability to think through more complex, multi-step problems – which is now also available to free-tier users. That opens new possibilities in training settings without requiring participants to pay for a subscription.
Of course, it’s not perfect. GPT-5 often responds more concisely than I’d like. The core is there, but sometimes the depth is lacking – especially in extensive analyses or strategic scenarios. The new, stricter usage limits are quickly felt by heavy users. And the “dynamic router” that governs how the model arrives at its answers makes the process somewhat less transparent – a small loss of control.
While GPT-5 isn’t flawless – context can still get lost in very long conversations, and particularly resource-heavy tasks like web or image processing can suffer delays – the positives clearly outweigh the drawbacks.
For my daily work, GPT-5 means more speed, greater reliability, and a communication style that feels more natural. The minor shortcomings are there, but they fade against the tangible gains. This new version is already a central part of my work – and I’m genuinely curious to see how it will evolve with future updates.
