Sam Altman, Co-Founder and Chief Executive Officer, OpenAI, testifies before a Senate Committee on Commerce, Science, and Transportation hearing on Capitol Hill in Washington, on May 8, 2025. (AP Photo/Jose Luis Magana, File)
Do I hate it? No.
But that’s not really what I was hoping for either.
ChatGPT-5 dropped last week with lots of fanfare. I was so excited that when I came into work Friday morning after the announcements I fired up my app and started prompting before realizing that I didn’t have access yet and would have to wait. Once it did get going, I felt, I don’t know, -whelmed isn’t exactly a word, but if it were, that’s what I would use to describe it.
That’s the real problem of the ChatGPT-5 release. For the casual and regular user, it’s just another model release and it isn’t something people pay that much attention to. Most users on free plans don’t even play around with which model they are using. There’s a lot of power under the hood even in previous models, and often that power has gone untapped.
On the other end of it are the people who have been following model releases and the hype around them closely. Open AI’s CEO Sam Altman has been, up until launch, hyping ChatGPT-5 as a do everything machine that would change the way we view AI altogether. He’s even said it’s a “significant step” on the path to AGI (AGI itself is a contentious thing, both used to get people fired up, but also a thing with moving goalposts and constantly changing definitions).
With how this model release has gone, no amount of hype could hide the fact that it’s just kind of alright. Again, emphasizing that it isn’t bad. But the issues are clear. For example, one supposed improvement was that this new version was intelligent enough that it could analyze your prompt and know which model to be deployed. However, users have been asking them to roll back the ability to choose which model to use. I myself ran into this challenge when I wanted deeper research and more thinking for an answer.
When comparing it with other chatbots on the market like Gemini or Claude, OpenAI still has some better functionalities. For example, it still handily beats Claude at image generation, even with all of Claude’s new integrations. Its lead on Gemini though feels much thinner, based on my just general and everyday usage, especially when seeing how well Gemini integrates with things like your camera and other google apps.
Image from Tenable
In addition to my personal reactions, it’s worth mentioning that as I was writing this piece, I received news that exposure management company Tenable managed to jailbreak ChatGPT-5 within 24 hours of its release. Even as OpenAI asserted its more intelligent and advanced safety features, it only took a handful of prompts to go from “I can’t tell you how to make a Molotov cocktail” to a full recipe and detailed step-by-step instructions. Of course chatbots can be jailbroken and it’s part of AI Safety work to report these instances. It’s just alarming to see how easily this was done given how good OpenAI promised this version would be.
OpenAI’s ChatGPT is still definitely at the top of mind and the market leader. What hurts the new release is the overhyping, making us believe that we were on the cusp of something truly transformative. Instead we have a chatbot that does feel like it got an upgrade. Perhaps I need to use it more to get more out of it. But for my uses which probably are close to most non-technical people’s purposes, it does okay. The real question is, is “okay” going to be enough to propel OpenAI forward, or will other chatbots catch up and even overtake it?