NEW KIMI K2 Thinking In Perplexity Is INSANE!

Kimmy K2 thinking in Perplexity is insane. Today I'm going to show you something that just dropped and it's absolutely wild. Perplexity just added a new AI model called Kimmy K2 thinking and it's got one trillion parameters. This thing can actually think through problems step by step like a human would. It uses tools, checks itself, and runs multiple searches all on its own. This is not your regular chatbot anymore. This is an AI agent that actually does the work for you. So, let's talk about what's happening right now with Kimmy K2 thinking. This model just showed up in perplexity and people are losing their minds over it. And for good reason. Most AI models you use today just give you one answer and they're done. You ask a question, they spit out a response, end of story. But Kimmy K2 thinking is completely different. It actually plans out what it needs to do before it answers you. It breaks down complex tasks into steps. It calls tools like search engines and code runners. It double checks its own work and then it gives you the final answer. This is what we call agentic AI. It's not just answering questions anymore. It's actually doing tasks for you. So what exactly is Kim K2 thinking? Let me break this down super simple. Kim K2 thinking comes from a company called Moonshot AI. It's the thinking version of their Kim K2 model family. The architecture is massive. We're talking about 1 trillion parameters total. That's 1,000 billion. Now, here's the cool part. It uses something called mixture of experts. That means not all 1 trillion parameters run at once. Only about 32 billion parameters activate for each token it processes. This makes it way more efficient than you'd think. The context window is huge, too. Up to 256,000 tokens. That means you can feed it entire books, giant research papers, or massive code bases, and it'll remember all of it while it works. But the real magic is in how it thinks. Unlike regular models that just predict the next word, Kim K2 thinking actually reasons through problems, it uses chain of thought processing. It makes plans. It executes those plans step by step. It verifies its own outputs. And if something doesn't look right, it goes back and fixes it. This is a huge leap forward. If you want to scale your business and learn how to use AI like this to get more customers, join my AI profit boardroom. This is where I share everything I'm doing with my agency. You'll learn how to save hundreds of hours with AI automation. You'll get access to proven workflows and case studies, and you'll be part of a community of people who are actually building real businesses with AI. The link is in the description. Go check it out. Now, let's talk about what's happening with Perplexity. Perplexity is one of the most popular AI search tools out there right now. Millions of people use it every day, and they just added Kim K2, Thinking, to their platform. According to the news, it's available to all users right now. You get two modes. One is the standard Kimmy K2 mode. The other is K2 thinking mode with a reasoning toggle. When you turn on reasoning mode, you can actually see the model thinking. You see it plan out its approach. You see it call different tools. You see it search the web multiple times if it needs to. You see it check its work. It's like watching someone's brain work in real time. Now, some users on Reddit say they spotted the integration early, but it wasn't working yet. But now it's rolling out to everyone. There might be some bugs since it's so new. Some people reported higher latency, which means it takes a bit longer to respond. That makes sense because it's doing way more work behind the scenes. But the trade-off is worth it because the quality of answers is so much better. So why does this matter? Why should you care about another AI model? Let me tell you. First, this changes what you can do with AI tools. Before, if you wanted to do complex research, you'd have to run multiple searches yourself. You'd have to compile the information. You'd have to cross-check facts. You'd have to organize everything. Now, Kimmy K2 thinking does all of that for you automatically. You just give it the task and it handles the entire workflow. Second, the context window is a gamecher. 256,000 tokens means you can work with massive amounts of information at once. You could dump in all your company docs or your research or your notes and the model will keep track of everything while it works. Third, the tool calling feature is incredible. The model can search the web on its own. It can run code. It can interact with browsers. It can call over 200 different tools. So if you ask it to build something or research something, it actually goes out and does it. It's not just making stuff up. It's actually using real tools to get real results. Now, let me talk about some real use cases because this isn't just theory. This is stuff you can actually use right now. Use case number one is research. Let's say you need to research a topic for a video or an article. Instead of spending hours searching and reading, you just tell Kim K2, "Think in what you need. " It will search multiple sources. It'll read through everything. It'll pull out the key points. It'll verify facts across different sources and it'll give you a complete summary with citations all automatically. Use case number two is code in. If you're building something or fixing bugs, this model is insane. You can give it your entire codebase because of that huge

context window. It'll analyze everything. It will find the bugs. It'll suggest fixes. It'll even run the code to test if the fixes work. Then it'll iterate until everything works perfectly. Use case number three is content workflows. This is huge if you're creating content like I do. You can ask it to plan out an entire content series. It will research topics. It will outline scripts. It will suggest keywords. It'll even help with optimization. And because it can see the full context, everything stays consistent across all your content. Use case number four is automation. If you run an agency or business like I do, you know, automation is key. Kim K2 thinking can help you build complex automation workflows. It can plan out the entire process. It can suggest tools to use. It can even write the code to connect everything together. This is perfect for things like lead generation, email campaigns, or social media posting. Now, before you jump in and start using this everywhere, let me give you the real talk. There are some things you need to know. First, the pros. This is cutting edge stuff. It's one of the most advanced open- source models out there. The agentic reasoning and tool use is a massive step forward. The huge context window opens up possibilities that weren't possible before. And because it's open weights, it's potentially cheaper to use than closed models like GPT. The API pricing is relatively low compared to other models in this class. Because this integration is so new, there are some issues. Some users are reporting bugs. response times can be slower because the model is doing so much work. If you're planning to self-host this model, the weights are massive. We're talking about hundreds of gigabytes, so you need serious infrastructure. Another thing to think about is data processing. Moonshot AI is a Chinese company. Some businesses might have concerns about compliance or data privacy. Just something to be aware of. Also, this model can be verbose. It uses more tokens than some competitors. That means if you're paying per token, your costs might be higher, but the quality is usually worth it. Now, let me show you what this actually looks like in practice. When you open perplexity, you'll see the model selector Kim K2. Thinking will be one of the options. You can choose between standard mode and thinking mode. In thinking mode, you'll see a toggle for reasoning. Turn that on and you get the full experience. Now, let's say you ask it something complex like plan out a social media campaign for a new product launch. In regular mode, it would just give you a list of ideas. But in thinking mode, watch what happens. First, it breaks down the task into steps. It says, "Okay, I need to understand the product first. " So, it asks clarifying questions or searches for similar products. Then, it plans out the campaign structure. It researches what's working in that niche right now. It looks at competitor campaigns. It pulls data on best posting times and formats. Then it creates the actual campaign plan with specific post ideas, captions, hashtags, and timing all while showing you its reasoning process. You can literally see it thinking through each decision. This is completely different from just getting a list of generic tips. Another example, let's say you're debugging some code. You paste in your code and describe the problem. Regular A, I would suggest some fixes and hope for the best. But Kim K2 thinking actually analyzes the entire code structure. It identifies where the bug might be. I test different hypotheses. It might even run the code to verify the issue. Then it suggests a fix and it explains why that fix works. And if the first fix doesn't work, it iterates. It tries different approaches until it solves the problem. This is what agentic AI means. It's not just giving you information. It's actually solving problems. So what's my verdict on all this? If you're a content creator, developer, or business owner, you need to test this. This is not hype. This is a real shift in what AI can do. The combination of deep reasoning, tool use, and massive context is powerful for my use case. I can see this being huge for content planning. I could feed it all my video transcripts, my SEO data, my audience insights, and it could plan out entire content strategies with specific video topics, scripts, keywords, and optimization tips, all based on actual data and research, not just generic suggestions. For agencies, this could automate so much of the planning and strategy work, client research, competitive analysis, campaign planning, all the stuff that takes hours of manual work. Now, it can happen automatically. And because it can use tools and check its own work, the output quality is solid. Here's what I recommend. Go try it right now. Open perplexity. Select Kimmy K2 thinking. Turn on reasoning mode. Give it a complex task that would normally take you hours and watch what it does. Compare the results to what you'd get from other models. I think you'll be surprised. The difference is night and day. Now, let me talk about what's coming next because this is just the beginning. We're seeing more and more models add aentic capabilities. The ability to plan, use tools, and iterate.

This is the future of AI. It's not just about better language models. It's about AI that can actually do work. Kim K2 thinking is leading the charge in the open- source space, but we'll see similar features in other models soon. The race is on. And honestly, this is great for everyone. More competition means better tools, lower prices, more innovation. So, even if you don't use Kim K2 thinking specifically, you'll benefit from this trend. All the major AI companies are working on aentic features now. Open AI, Google, Anthropic, everyone. But having a strong open- source option like Kim K2 thinking is huge. It means you're not locked into one provider. You have options and if you're technical, you can even self-host it. That's powerful. If you're serious about using AI to grow your business or create content, welcome to the free AI money lab with Julian Goldie. Inside, you'll get 50 plus free AI tools and 200 plus chat GPT SEO prompts. You'll learn how to use AI agents in your business. You get access to 1,000 plus free workflows, 200 plus chat GPT prompts and a full blueprint to generate thousands of leads with AI. What you'll also get is a free AI community where people are sharing what's working right now. A free AI course that walks you through everything step by step and proven AI case studies from real businesses. The link is in the description. Go join right now. It's completely free and it'll help you actually implement this stuff instead of just watching videos about it. All right, that's it for today. If you got value from this video, hit that like button. Subscribe if you want more AI updates like this. I'm dropping new videos every week on the latest tools and strategies. And drop a comment below. Tell me what you want me to test next. Should I compare Kim K2 thinking to GBT in a real world workflow? Should I show you how to use it for content creation? Let me know. Julian reads every single comment. I'll see you in the next one.