Please note javascript is required for full website functionality.

Blog

AI Blog: OpenAI ChatGPT-5

15 August 2025

Welcome back to our AI blog series!  You may have noticed that OpenAI has officially released GPT-5 on August 7, 2025.  It is expected to be a milestone that signals both technological advancement and evolving user expectations.

 

As OpenAI introduced, GPT‑5 is not just an upgrade but a reimagined system.  All the prior models including GPT‑4o, GPT‑4.5, GPT‑4.1, various mini and o‑series models have been merged into one model that dynamically routes tasks between fast responses and deeper thinking modes.  

GPT-5 boasts great performance in the following aspects:

  • coding and reasoning prowess: scoring 74.9% on SWE-bench Verified, 88% on Aider Polyglot and delivering 94.6% on the AIME 2025 math benchmark without tools.  It achieved an even higher rate of 88.4% on GPQA under the extended thinking mode
  • hallucination reduction: according to OpenAI’s tests, “GPT‑5 thinking” shows a sharp drop in hallucinations, with approximately 45% fewer factual errors than GPT‑4o in “auto” mode and an error rate reduction of circa 80% compared to earlier o‑series models.    
  • multimodal excellence: GPT-5 has a stronger performance across vision, video, spatial and scientific reasoning tasks.

In some specific domains like medicine, GPT‑5’s zero‑shot, multimodal reasoning outperforms human experts on key benchmarks, showing its strong capability in understanding and reasoning.

As OpenAI has emphasised, GPT-5 can bring tangible improvements to everyday applications.  It can be your writing assistant which is always online and ready to improve your expressions.  It can help with health inquiries when you ask medically relevant questions.  Compared with prior models such as GPT-4o, GPT-5 is more adept at managing multi-step tasks, chaining calls across tools and adapting to shifting contexts.

Although many users of GPT-5 have provided positive feedback, this new model has been criticised for the shorter, less creative responses, a robotic tone and limit-driven usage caps.  Some users complained that “…answers generated by GPT-5 are shorter and, so far, not any better than previous models.  Combine that with more restrictive usage, and it feels like a downgrade branded as the new hotness”.

In areas like coding, healthcare and legal work, GPT-5 is raising the standard for how smart, flexible, and reliable AI can be.  Its new design makes it easier to use and can think more deeply when needed.

However, the launch also showed something important: people don’t just want a smart AI but also want one that feels human.  Things like warmth, personality and being predictable are as important as raw performance.

The release of GPT-5 is a reminder that great technology needs to connect with people emotionally too.

 

Join us next time for more exciting news of AI tools!

Newsletter