Ladies and Gentleman there cannot be a larger buy signal than this. The fact that this model requires so much power it has to be limited to 35 messages per week is incredible. I am driving this one point because it is a clear throughput situation.

Also, the information cutoff is October 2023 which tells me this isn't the new massive model they are releasing. So, what does this all mean?

It means they released a new capability which uses much more compute but isn't technically perhaps their new monster model that is trained on trillions of parameters. If this was a new model I would expect the cutoff date to be much more recent. At least 3-6 months would have made more sense. Now, there is time between training and cutoff time but I don't think they've been sitting with a model that has been trained for over a year ago. A newer, more recent model must be sitting there for use but can't.

What this all suggest to me is that they are currently massively GPU constrained at the moment and waiting desperately for Blackwell. What you would expect is massively more throughput for GPT-o1 and you eventually have to expect a newer more massive model that is fundamentally better at the core.

Joanne Jang from OpenAI tweeted this.

"There's a lot of o1 hype on my feed, so I'm worried that it might be setting the wrong expectations. what o1 is: the first reasoning model that shines in really hard tasks, and it'll only get better. (I'm personally psyched about the model's potential & trajectory!) what o1 isn't (yet!): a miracle model that does everything better than previous models. you might be disappointed if this is your expectation for today's launch—but we're working to get there!"

OpenAI also said they are adding additional tools later.

So in effect, they released a new engineering marvel with model actionability/capabilities i.e. thinking - in preparation for adding a massively more capable model new thinking capability later. It's going to be a 1, 2 punch that I don't think the world is ready for.

The 35 messages per week is a throughput drop I could not have imagined. This is what Jensen was talking about that people are getting emotional.

The amount of GPU spend from the new AGI race that is about to ensue is going to be insane.

Positions - Super Long Nvidia with a position in ARM, AVGO and Microsoft