Beth Lyons

I thought it was 3.5, but it might've been 3.5.

So over the weekend, this actually, this news story comes from Carl in the ghost that he is right now, but he posted on our private Slack.

But over the weekend, DeepSeek launched three different models.

I think all Reasoning First built for agents.

So DeepSeek version 3.2 and DeepSeek version 3.2

I am saying that that's how you would pronounce that word.

So the Reasoning First model is built for agents.

3.2 is the official successor to the experimental model of the same version that was released earlier.

It's live on the app, web, and API.

I'm sure you can play with it on OpenRouter if you want.

And special is is ABI only for now.

And that is I don't see more information about what that is.

What makes it special or special?

So I think in terms of what you're saying, Brian, Deep Seek has has thrown the first gauntlet top that in the December death match or, you know.

Clearly, they didn't take from Wednesday off, right?

So, because it's not a holiday over there.

So, like, it was just a day.

We can't hear you.

Like weeks ago.

Go for it.