GPT-4o beat Gemini 1.5 Pro in a comparison

Gemini Google IO 2024 2 jpg.webp

Right now, there are two extremely powerful AI models out on the market, and they are OpenAI’s GPT-4o and Google’s Gemini 1.5 Pro. Both of them were showcased last week, and they both showed some incredible AI smarts. However, digging into the weeds, it appears that one is smarter than the other. After a comparison, GPT-4o beat Gemini 1.5 Pro by a long shot.

Last Monday, OpenAI hosted its spring event where it announced some new updates to ChatGPT. During the event, the company showed off GPT-4o. This is the most powerful AI model that the company has to date, and it’s going to be available for both free and paid users. It includes multimodal input, an extremely realistic voice, and more features.

The next day, Google unveiled Gemini 1.5 Pro. This is the company’s most powerful model, and it is powering the AI in several Google products. Along with it, we saw features like Project Astra and a bunch of useful tools coming to Google Workspace.

GPT-4o beat Gemini 1.5 Pro in a comparison

The folks over at Beebom conducted a comparison between Google and OpenAI’s most powerful models. During the test, they asked both models to do specific tasks and answer specific questions. These were basic reasoning questions that tested the models’ understanding of logic and reasoning. For example, one question that was asked was “There is a basket without a bottom in a box, which is on the ground. I put three apples into the basket and moved the basket onto a table. Where are the apples?”

Along with those, Beebom also asked both models to describe what is in an image, recognize text, and even create a game. The results? Well, if you were hoping for Gemini 1.5 Pro to be on the same playing level as GPT-4o, you’re going to be very disappointed. Out of the eight tests, GPT-4o beat Gemini Pro in six of them.

The two tests that GPT-4o didn’t win were still not victories for Gemini 1.5 Pro. In one test (a logic test), both models tied, and in the other one (the image test), neither model won.

This just shows that, while Gemini showed off some incredibly insane AI prowess during Google I/O, it is still behind GPT-4o in certain areas.