GPT-4
OpenAI's 2023 model, the one that made large language models mainstream and held the top spot longer than anything since.
GPT-4 arrived in March 2023 and reset what people expected from a chatbot. It passed exams, wrote usable code, and was accurate enough that the mistakes became interesting rather than constant. Every model since has been measured against it in some way.
Two details make it a useful reference point. It stayed at the top of capability rankings for about a year, whereas leaders since then have typically held the position for around seven weeks before someone passes them. And it was trained on text alone, unlike today’s models, which also learn from images, audio and video. When people are surprised that GPT-4 could sketch a passable unicorn in code without ever having seen a picture, that is why.
-
A Robot That Only Tracks Objects Predicts the Wrong Move. Somebody Finally Measured By How Much
-
Someone Else Is Paying for Your AI, Just Not the Part You Think
-
Stop Asking Whether AI Is a Bubble. Start Reading the Lease Agreements.
-
Karpathy turned one paragraph of Tolkien into a 3D scene for ten dollars, and called it a vibe check