Open AI Model Confusion!
May 15, 2025
A few quick notes from the intersection of AI, marketing, branding, and small business.
Not sure it could be more confusing. For all the smarts that Open AI seems to have in-house, you’d think they could do a better job naming their products. Even people who exclusively use ChatGPT every day have trouble understanding the differences in their six models. And what about people who use ChatGPT as well as Grok or Claude or Perplexity? No chance! BTW, ChatGPT supposedly has 800 million daily users, so that’s a lot of people to confuse. In my opinion.
It’s worth figuring out. The fact is, the various Open AI models DO have significant differences, and while it’s annoying they don’t do a better job at naming, it makes a difference which one you use. If GPT 4o is all you ever work with, you’re probably getting substandard output, especially for complex tasks or externally facing communications where brand voice and EQ matter. For small businesses interested in work like reports, market research, brand communications, marketing, and such, three of Open AI’s six models are the ones to focus on. For a look at all six of their models, check out the May 12 edition of The AI Daily Brief podcast. Super helpful. Or refer to this Open AI model cheat sheet put together by Charlie Hills.
Open AI’s GPT - 4o. This is currently (at the current pace of change, this post will probably be relevant for about 3 weeks), the default model, and the one most people are using. Open AI says it’s great for everyday tasks. Honestly most of the time 4o works so well that you might wonder why there are other models at all. You get unlimited queries and it’s fully multimodal, which means it can handle a whole bunch of file formats and create images.
Open AI’s GPT - 4.5. This model is the best for creative tasks that involve external communications and writing that requires high empathy and EQ. This is probably the most nuanced writer of the Open AI models. However, you’re currently capped at the number of queries per week, so use 4.5 judiciously.
Open AI o3. Given the name, you might expect that this is an inferior model to 4o or 4.5, but you’d be wrong. o3 is your strategic thinker and deep researcher. It can be a true thought partner, while the other models are more like assistants. Use o3 for strategic plans and risk analysis. BTW: whenever you select “deep research” on a ChatGPT query, you’re using o3, which is just another Open AI way of confusing us! 03 also has a weekly usage cap, but it’s not as restrictive as 4.5.
Better results. So in the same way that careful prompting will lead to much better results, using the right Open AI model will give you better output as well. And of course you’ll get even better outputs if you’re using a Brand Gravity guide to inform the model’s thinking.
Not so fast. All that said, yesterday’s The Most Interesting Thing In AI podcast definitely throws some shade on Open AI model advancements, and really the whole current AI hype. Gary Marcus says that GPT - 4.5, which was slated to be GPT - 5.0, isn’t really all that much better than 4o and that the insane progress over the past few years has already slowed considerably. The show is definitely worth a listen.
I’ll get into the different models from other AI companies in future posts.
