Keri Briske

And that's honestly what we've been trying to deliver with all these efficiencies that we've been doing.

And then the super kind of gives that

more accurate than the Nano, not quite as accurate as the Ultra.

Actually, our latest super revision just surpassed our Ultra.

So we've got our roadmap going.

There's always innovations happening.

But the Ultra does have safety and robustness tests and capacity to learn.

So a lot of, like I've talked to healthcare customers who definitely prefer the larger model because of its capacity to learn on their domain and be more robust for a given task.

So I hate to say it, but it depends.

Normally, they're going for more accuracy or they've acquired more data in what we call the data flywheel.

So the second you put a model out, and actually sometimes it goes in reverse.

So you put a more accurate model out, the bigger model, and then once you get your data, then you start to be able to fine tune and then put the smaller model out.

So it almost sometimes goes in reverse.

But for people that are just starting out with the nano and want to go larger, it's really about understanding your use case

understanding how to evaluate it.

Most businesses actually kind of get stuck in this evaluation phase.

It's really hard.

When you start a proof of concept, you have to really narrow down the scope.

It's like any other problem.

How will you test it?