Andy Halliday

Yeah.

Yeah.

So that's the strategy that I think is going to work for Amazon.

But here's how good Nova 2 Pro is, and it's just been released.

So it's equal to or better on 10 of 16 benchmarks compared to Claude Sonnet 4.5.

Now, that's not compared to Claude Opus.

Yeah, but still.

Yeah.

Sarnit 4.5 is a model that everybody ought to be thinking about using because it's jammed near as good in every way.

It's equal or better on 8 out of 16 benchmarks compared to GPT 5.1.

And it's equal or better on 15 out of 19 benchmarks compared to Gemini 2.5 Pro.

Yeah.

Which was a darn good model.

It was, yeah.

Oh, by the way, it's equal to or better on eight out of the 18 benchmarks comparing it to Gemini 3 Pro.

So it's not hitting the top marks in reasoning, but a lot of the things that are really useful to development of actual working applications, it's right there.

It's equal to or better than in those cases.

So that's the NOVA line of models.

And so then there's one other thing I wanted to mention about what they announced here.

I mentioned Nova Forge, which is the system that allows customers to make their own Frontier models, you know, kind of bespoke models that way.