Beth Lyons

Uh, and then you have Claude Opus 4.6 max, just one point ahead of that.

So bracketed around news.

So it's a definitely, um, it's, they've staked a claim in the closed frontier model space.

And that really is a tremendous leap from look way over here on the right, where Lama for mad Maverick was at 18.

It barely stayed on the chart.

Um,

But now look at what follows that.

So there's news about GLM 5.1, this one at 51.

And then that was just released.

It's a 744 billion parameter model, 40 billion active.

It tops SweeBench Pro and hit number one on coding benchmarks ahead of GPT 5.4, not in overall intelligence, but in coding.

That open source model is definitely a leader.

And then you see a host of open source models, including Quen 3.6+, Minimax.

And then down here is Grok 4.2.

This is why XAI is trying to build multi-trillion dollar models now to try to get back up to the frontier level.

And they will.

I mean, they're definitely going to use scaling principally as a way of achieving

parity and perhaps exceeding where Gemini 3.1 is right now, but don't count Google out.

They'll jump ahead as well.

But now let's talk about open source because Carl is using open source and increasingly that will be the way to go.