In today’s episode, I’m providing you with the five most important takeaways from the release of OpenAI’s long-anticipated GPT-5 model.
My first big takeaway is that, unlike the leap from GPT-3 to GPT-4, the transition from GPT-4 to GPT-5 may not feel as groundbreaking and certainly a lot of folks out there have been expressing underwhelm about the model. This underwhelm, however, is misplaced. As evaluations by METR (Model Evaluation and Threat Research) cleanly illustrate, GPT-5 performs about where you’d expect the world’s leading LLM to perform (or even a little better than expected).
Second, GPT-5 consolidates several different LLM capabilities into a single model experience. Prior to GPT-5’s release on August 7th, you might use GPT-4o if you were prioritising speed, GPT-4.5 for high-quality creative writing and o3-pro for challenging mathematical or coding tasks. Like I’ve become accustomed to with Claude Sonnet 4 and Opus 4 for several months now, GPT-5 automatically determines how much behind-the-scenes processing (“reasoning”) it should do before beginning to output a response. This is convenient for sure but, in this case, OpenAI is catching up to Anthropic instead of being the leader.
Third, since the advent of LLMs, naysayers have been complaining that because of hallucinations, generative and agentic AI applications have limited viability in serious commercial or industrial use cases. I’ve already found that the hallucination rates since GPT-4, but particularly since agentic approaches like Deep Research were released, are nearly negligible. Well, GPT-5 makes big big strides here.
Fourth, as I reported on in detail in Episode #908, LLMs are prone to dangerous deception especially when their objectives are threatened. Like hallucinations, this is another area where GPT-5 makes huge strides, making GPT-5 much safer to use within agentic applications than OpenAI’s predecessor models.
Fifth and finally, as the great Dr. Andriy Burkov recently pointed out in a viral LinkedIn post, with GPT-5 performing only on par with other existing proprietary models such as Claude Opus 4 on key benchmarks like SWE-Bench, the time to get super-excited about what the next cutting-edge LLM will be able to do is past. Now is the time to get super-excited about what you can building and accomplishing with LLMs. The tools available to AI practitioners are extraordinary. What process can you automate to a high degree of accuracy? What new capability can you improve society with? Not sure… as I’ve said so many times on this show, an LLM conversation to ideate on what you could be doing with cutting-edge AI tech is merely a browser click away.
The SuperDataScience podcast is available on all major podcasting platforms, YouTube, and at SuperDataScience.com.