Beth

So a player in the marketplace.

And it says it's optimized for enterprise and can run on as few as eight NVIDIA GPUs, which is also pretty decent.

Absolutely.

And they're designed for edge devices like your phone, offline use, customization.

And critically, the promo is 10x fewer tokens generated than competitors.

And tokens are where the money comes from, right?

Tokens are where you start getting charged for usage.

So, yeah, very exciting.

I will also say we know a little bit more about OpenAI's models that are coming out a little bit more released today.

It looks, to recap what we said before,

OpenAI, Sam Altman, told employees the company is moving into a code red surge to improve ChatGPT after Google's recent updates.

And that, of course, is reminiscent of Google's own code red when ChatGPT came out and took Bard to the mat.

So there's a model they're developing codenamed Garlic.

And that garlic is a larger model upgrade and that's coming out next year.

Basically, the information says it could be a 5.2 or a 5.5 type release.

So it's not going to be a new number.

But there is also a new reasoning model supposedly launching next week.

that reportedly beats Gemini 3 on benchmarks.

Now, I find this news a little confusing.

Why would you release a new reasoning model that doesn't seem to be an iteration of the five next week and then release an iteration of the five in 2026?