Is Ramp an AI Company Now? Plus Kimi K3 Beats Fable on Front-End | This Week In AI

**Shane Thomas** (0:00)
Gemini and Google have been sleeping a bit at the wheel, it seems. So they released three new Gemini models.

**Abhi Aiyer** (0:05)
They're like a dark horse in this race, but they're just not doing what we expected them to.

**Shane Thomas** (0:10)
They're like a sick dark horse. They're like in the back, they're like struggling to keep up. They're still there.
Thank you all for tuning in. We're live in London.

**Abhi Aiyer** (0:30)
In London, mate.

**Shane Thomas** (0:31)
We are having an offsite this week with Mastra, so we're hanging out with a lot of the team members. But like we do every week, we're going to do the news. And normally we do this on Monday, but because of the travel, because of the conference week, we're doing it on Tuesday. So if you're tuning in live, thank you. If you're watching after, go give us that review.

**Abhi Aiyer** (0:48)
Is it a review again?

**Shane Thomas** (0:49)
Because you should. Give us that five star. We appreciate it. It helps other people find the show. All right. We are going to go through the news for this week. We're going to do it pretty quick. There's a few big things, but it wasn't a crazy news week. I think there was one big model drop, and that's where we're going to spend most of our time. Kimi K3.

**Abhi Aiyer** (1:06)
Kimi K3.

**Shane Thomas** (1:07)
We've got to talk about it. So this was on July 16th. Introducing Kimi K3, Open Frontier Intelligence. It's a 2.8 trillion parameter model, a million token context window, native multimodal, and then let's look at the benchmark. So this one is across knowledge work and agentic browsing. You can see Kimi K3 is right there with Fable 5, which is, you look at, what are the top models? It's Fable 5 and GPT-56, right? That's what everyone's comparing against. And the general consensus is Fable 5 is maybe slightly better at quite a bit of things than 5, 6, but not a lot better. 5, 6 is just right there. And then 4, 8 is maybe around 5, 6, but maybe just below. I think that's, is that how you would roughly?

**Abhi Aiyer** (1:50)
I would say, and then there's a big price difference between these things as well.

**Shane Thomas** (1:54)
Yes, Fable being very expensive. But Kimi K3 is kind of right there. It beats Fable in some things, it beats Sol in some things, and again, this is just knowledge work and agentic browsing.

**Abhi Aiyer** (2:04)
And cheaper price, which is wild.

**Shane Thomas** (2:06)
But it isn't open model cheap the way you'd expect. They kind of priced it to be comparable. I think it's still cheaper than OpenAI models, but not that much cheaper.

**Abhi Aiyer** (2:15)
Yeah, it's like Sonnet level prices.

**Shane Thomas** (2:16)
Yeah, it's like Sonnet prices, which most of the Chinese models, they haven't priced it that way before. So it's expensive when you consider, compared to most open source models, more open way models.

**Abhi Aiyer** (2:26)
And their hardware isn't as good. So I mean, they're going to be working on it, of course, but you're paying cheaper. Maybe the performance is whatever now, but you get the results.

**Shane Thomas** (2:34)
Yeah, the results are pretty close to frontier. And so Kimi K3 scores 57 on the Artificial Analysis Index. It's comparable to Opus 4.8 and 5.5, but it's behind Fable and 5.6. They were going to release the open weights, which would make it the leading open source model. And it's not really that close, right? It's quite a bit higher than all the other open models. Then I started seeing some additional things that came in, which was very interesting. So big news, Kimi K3 is now number one on the front end code arena with 1,679 points surpassing Fable 5
So this is like front end coding, right? And then it's visually number one on the front end web app arena by Design Arena. And so it beats Fable 5, Sonnet 5, Opus 4-8.

**Abhi Aiyer** (3:18)
I think it's very interesting because when we talk to Marvin and Damien, our front end engineers, they always comment on how bad the models are at front end in design in general.

**Shane Thomas** (3:28)
It's like you can make, you can kind of follow design sometimes, but it writes, as Marvin would say, really shitty react code. You can get the job done, but it's not great at it.

**Abhi Aiyer** (3:37)
I think it's so interesting that Kimi is really good at front end because that urges a lot more people to use it.