Kimi 2.7 Code: This Model Built THIS in 15 Minutes ($15/M)

The new Kimi K.27 is here. In today's video, I'm going to be testing it out. I'm going to give my opinion, etc., etc. I'm not going to waste your time. Let's just jump into things. Okay, so I'm just going to paste the prompt. This is the school community prompt that I use at the beginning of every new model just to see how it does with this prompt. Now, the things I'm looking for are speed, right? The way, and also uh just yeah, how good the model actually is. So, I'm This is Kimi code, if you didn't know, which I do actually quite like Kimi code. I don't use it per se, but I will say that when I do use it, I do like it a lot. Now, Kimi is known to me as an intelligent model, a fast model, and actually quite a good model as well. So, I will just start with that just so you guys know. I think that this is going to be a good model, but we'll see how it does, and we'll see how close it is to actually replacing something like Opus. Now, another thing that Kimi is known for is that it's like a cheap model, but also a very efficient model. So, it's actually slightly more expensive than Kimi 2.6. I just checked, and it still has the 260,000 contacts window. But, what I will say is this model is known for being extremely efficient and also pretty fast, too. So, right now, it's just kind of getting on with things right now, so we'll let it run, and we'll see how it does with this task. Just so you guys know, the kind of criteria for this task is like font choice, what it looks like, if there's any technical problems with the website, etc., etc. The prompt, basically what it should do is it should make a Next.js service-based website for a made-up business. We've got context here, and also we've got how much we've actually used. If I go here, and I can refresh this, for example, and we should see a bit of usage coming in here. So, yeah, I've I've used nothing so far. And we'll see how it this goes. Just before we continue with the video, guys, huge shout-out from our sponsor, me. If you don't know what Harbor is, Harbor is my baby. This is the tool that I have created from scratch and that basically you can go and get today. I've brought back founder pricing for another 18 days, guys. You have exactly 18 days to lock in founder pricing. What does founder pricing do? It basically gives you 35 articles instead of 25, 75 articles instead of 50, and 150 articles instead of 100. So, it's about What's that? Like 50% more articles, I think it is. Mathing Math ain't mathing right now. And if you're curious about the results of Harpa, it's an SEO content generator, but it's also just an AEO SEO tool as well. We've published 1,848 pages. We have almost 600,000 impressions and 9,500 clicks. These are all real figures from real businesses. We have about 100 search consoles connected right now and that data is from those search consoles. Go and check out Harpa today, guys. There's a link in the description and in the pinned comment. I have huge ideas for this tool. We just released dev, which means you can use this directly inside Claude Code to write amazing content that ranks on Google, ranks on Bing, ranks in ChatGPT, ranks everywhere automatically using either the MCP with Claude Code or just the API. Go and check out Harpa today, guys. Link in the description, link in the pinned comment. It's the best way to support me as well. Let's jump back into the content. So, yeah, just So, everyone is on the same page here. This is kind of what I'm waiting for. I'm waiting for a model like Moonshot Kimmy K2.7 code and I'm not saying it's this model, but that I can use inside Claude Code at the same level as, you know, Opus, Fable, whatever. I've said from the beginning Fable for me, Claude Fable 5, is not that It's not that important of a model for me because I can I was already doing everything that I'm doing now with Opus 4.8, right? So, there gets to a time where the the gap between open source and closed source models gets to the point where it's no longer needed to pay Anthropic an absolute crap ton of money every month. And that's what I'm waiting for. Not really for me, but to tell you guys, right? Because I know that not everyone wants to pay. I mean, we pay two grand a month to Anthropic for their API, right? To power Harbor, Claude Code. We have two max plans, you know, there's a lot going on. We pay a lot of money to Anthropic. For us, it's fine because we're a business. I say we, you know, I have a business with a business partner, right? So, we are fine spending that amount of money, but I know that not everyone is. So, like what I'm waiting for is for the day that I can say, go and use this model inside Claude Code. Just so you guys know, Claude Code router, it allows you to use the power of Claude Code with any model, right? So, I could be using Kimmy 2.7 inside Claude Code right now with this model. Another one is Open Code, same concept, right? It's like Claude Code. I personally would rather use Claude Code, but Open Code is also very very good in my experience, and I'll be testing this with Kimmy 2.7 probably tomorrow. Okay, guys. So, that is officially the coding done, right? And I want to draw your attention to a few things. So, first of all, despite people thinking that 1 million context is like the the minimum, 262 works fine as long as the model is efficient with that usage. That's my cat meowing. Sorry, guys. What is it? I know. And then the other thing I want to point out is this would have been extremely cheap had I paid for the API costs. So, just bear that in mind when we talk about kind of the next steps and what it looks like, etc. This is possibly the most efficient model, right? And if you're really struggling for budget, this is probably the model I would be using. Okay, so let's have a look at the final result. Okay, this is definitely an interesting one. Damn, those SVGs are really nice. I have to say this is some of the best SVG work I've ever seen. The Chinese models are good at SVG, I have to say. It's very It looks very ChatGPT-esque, um, which is very interesting. Okay, well that just died on me, interesting. Okay, so the main thing here is making sure that the technical build is correct. So yes, you can swap to English. You can click on company Omni. Wow. That's That SVG is crazy good. That is probably one of the best SVGs I've ever seen. Wow. That Okay. Uh, this is a very good technical build. There's not a single issue here. The SVGs are next level. They are some of the best I've ever seen by far. The quality here is very, very good, I have to say. Let's see if I swap to English. Wow, it even keeps the page. This is only something that the really best models will do. Normally what happens is either A, the language switcher is broken, or B, when you click it, it goes back to the home page in the new language. It's very rare that this actually works on the page. Crazy. Get a quote, beautiful. I mean this is the best open source model I've ever seen by far. Kimmi does cook, I have to say. I've said this over and over and over, Kimmi cooks. And yeah, I would I I would really like to see the price on this. I can check my usage, but that's not quite the same thing. So looks like I used 40% of my quota. Which is actually quite a lot, but I am only paying 15 a month, and 8% of my weekly quota. So technically, math in my head, I could do about 12 of these websites every week for $15, which is probably a lot more than Claude would give me, to be fair. This is impressive, guys. I have to say the quality is um I think I'm I think it's easily the best I've ever seen from an open-source model, easily. Let's just see if this splits properly into individual locations. It does. It's done all the SEO. Literally SVGs are beautiful. The look and feel is phenomenal. It does kind of look ChatGPT-E, but it it it is really really nice. Guys, I think this is one to watch. This is what I would be using right now if I didn't have the money to pay for Claude. 100% I'd be using Kimmy 2.7 inside Kimmy Code. 15 bucks, you get pretty damn good usage, I have to say. Maybe not the 5-hour usage. I think that could be improved. MiniMax is definitely the most efficient model, but this is um yeah, this is very very good. This gets a really high score for me. This is like a 9 out of 10, honestly. The technical build is amazing. It really is. Anyway, guys, I could talk about this all night, but this is very very good. Very impressed. Shout out Kimmy. Like I said, I do like Kimmy. I think they're cooking over there. Honestly, I don't know what they're cooking, but they're cooking. Um and I hope they continue cooking because that is something else. Look at that. That is so good. Wow. Thank you so much for watching, guys. If you are watching all the way to the end of the video, you're an absolute legend. Go and check out Harbor SEO.ai. I will leave a link in the description. And as usual, peace out.