OpenAI’s o1 vs GPT-4o vs Perplexity: Fact-Checking = Sources > Reasoning

Leonardo.ai + Infogram

In my previous exploration, I pitted OpenAI’s o1 model against Sonnet 3.5 in a coding challenge. The o1 model impressed me with its ability to tackle complex tasks where Sonnet 3.5 struggled. Building on that excitement, I wanted to dive deeper into the capabilities of these AI models, but this time in the realm of fact-checking and internet research.

But o1 does not have an internet search, so how?

The Contenders: OpenAI o1, GPT-4o, Custom GPTs, and Perplexity

For this experiment, I employed:

My CustomGPT I use for quick fact checking

Do not have a Medium Account? Read here!

• OpenAI’s o1-preview model: hyped to be a breakthrough in reasoning, but without browsing/code interpreter and other tools access yet
• GPT-4o: A faster, cheaper, and more accessible model with internet browsing and other tools access
• Custom GPTs: Specifically prompted for fact-checking GPT-4odesigned for specific tasks.
• Perplexity: A startup focused on…