The Verge says GPT-5 is coming in December.
This is big news—not surprising for those of you who dared read the tome I wrote before Summer, where I said the most probable release date for GPT-5 was right after the US election, but it is big news nonetheless.
OpenAI CEO Sam Altman says this is “fake news”.
The company contacted The Verge to tell them they’re not releasing anything named “Orion” this year (Orion is the internal codename for the thing everyone calls GPT-5). In a Reddit AMA, Altman and others from OpenAI emphasized that “nothing” called “GPT-5” is coming out this year.
But we don’t care about what OpenAI says. They’d dismiss this as false even if it’s true. Altman doesn’t want Google to do to him what he does to them—steal the spotlight.
Just to play it safe, I’ll stick to The Verge’s reporting, as Altman has more reasons to blur the truth, and if, by chance, he’s being forthright this time, The Verge’s account won’t miss the mark by much.
Anyway, schedules aside, whenever Orion/GPT-5 arrives, the entire world will be waiting with anxious, hopeful, or dismissive curiosity for the first reactions from analysts and users.
Given the recent streak of bad news in the generative AI industry (including but not limited to bad training runs that are delaying—or right out halting—the next gen of models), GPT-5 will make or break our expectations for the coming years.
So, taking advantage of the news, I’ve decided to write another article about GPT-5. One that’s nothing like the 14,000-word monster I published before summer.
If that one was a detailed and labyrinthine treatise, this one is a quick and straightforward recipe.
The ingredients are information, rumors, knowledge, predictions, expectations, etc. that I’ve compiled below into brief sazen-like bullet points so you have the entire picture.
Take away what you will.
Without further ado, or whatever this was, here are 15 things I think you need to know about Orion/GPT-5.