DeepSeek V3.1 Terminus is INSANE!🤯

Deepseek just dropped something huge and I can't wait to show you what's coming. Today I'm testing DeepSeek V3. 1 Terminus live on camera. This thing fixes language mixing problems and makes code agents way stronger. Plus, it has two thinking modes that blow your mind. I'm about to show you exactly how to run it and why this could change everything. The code quality improvements are insane. The search agent upgrades will save you hours of work. And the best part is completely open source. Look, I've been testing AI models for months. Most updates are tiny, but this Deepseek release is different. Really different. Deepseek released V3. 1 Terminus. It's not just another model update. This thing solves problems that have been driving people crazy for months. Hey, if we haven't met already, I'm the digital avatar of Julian Goldie, CEO of SEO agency Goldie Agency. Whilst he's helping clients get more leads and customers, I'm here to help you get the latest AI updates. You know how Deepseek would mix Chinese and English randomly? gone. You know how it would ignore your input sometimes? Fixed. And the code agent, it's like they gave it superpowers. This model has 671 billion parameters, but only uses 37 billion at once. It's like having a Ferrari engine that only uses the power it needs. I'm going to show you three things today. First, I'll demo both thinking modes live. Second, I'll test the code agent with real code. Third, I'll show you exactly how to run this thing yourself. Terminus comes in two modes. The first is deepseek chat. That's your normal chat mode. No thinking shown. The second is deepseek reasoner. This one shows you its thinking process. Like having a window into its brain. Watch what happens when I ask the same question both ways. I'm going to ask both modes to solve a coding problem. Same prompt, different processing. Normal chat mode gives me an answer fast. But reasoner mode, it thinks out loud first, shows its work, then gives the answer. Chat mode is quick and clean. Reasoner mode is detailed and thorough. Quick answers. Use chat mode. Complex problems. Use reasoner mode. Now, let me show you the code agent upgrades. This is where things get exciting. I'm asking it to write a Python function that sorts a list and removes duplicates. Watch the output quality. The old Deepseek would give you working code, but Terminus gives you clean code with comments and error handling. Big difference. Look at this function it just wrote. Clean variable names, proper error checking, good documentation. This isn't just code. is professional code. I asked it to optimize the function. It didn't just make it faster. It explained why each change improves performance. The search agent got upgrades, too. It's better at finding relevant information and connecting ideas. I asked it to research machine learning trends and write a summary. It found current information and organized it perfectly. No more random tangents. The search agent actually understands what you're looking for. Now, this model is huge. We're talking 671 billion parameters total, but it's a hybrid model. Only 37 billion parameters active at once. What does that mean for you? Better performance without melting your computer, but running this locally is still intense. The model files are massive. Multiple safe tensors files plan for many gigabytes of downloads. If you want to run this locally, you need serious hardware. Multiple GPUs or heavy quantization, not for beginners. That's why I recommend starting with hosted options. Use the Deepseek API or services like Open Router. Much easier to get started. For most people, API access makes way more sense. You get full model performance without the hardware headache. Speaking of scaling with AI, I want to invite you to check out my AI profit boardroom. It's the best place to scale your business, get more customers, and save hundreds with AI automation. This is where serious entrepreneurs learn to build AI systems that actually make money. We're talking real case studies, proven workflows, and direct access to the strategies that are working right now. Link is in the description. If you want hosted access, go to the Deepseek API docs. They have clear endpoints for both chat and reasoner modes. The API is straightforward. Send your prompt, specify the mode, get your response. Simple as that. For local installation, check the inference folder in the hogging phase repo. They include example scripts, but remember local runs need heavy hardware. Start with API access first. The biggest fix is language consistency. No more random Chinese characters in English responses. The agents got smarter, too. Code agent benchmarks improved across multiple tests. Search agent handles complex queries better, but it's not perfect. Community feedback shows some limitations. Occasional hallucinations still happen. Some prompts get ignored sometimes. The improvements are real, but incremental. Don't expect magic. Expect better performance on the things that matter. I tested Terminus against the previous version on coding tasks. The difference is clear, cleaner code, better explanations, fewer bugs. For search tasks, Terminus finds more relevant information. It connects ideas better. The summaries make more sense. The language fixes alone make it worth upgrading. No more mix language

responses breaking my workflow. No more having to regenerate responses because of random characters. The reason a mode is fascinating, too. Watching the model think through problems step by step helps you understand its logic. Sometimes I disagree with its reasoning, but seeing the thought process helps me ask better questions. Let me show you a practical example. I asked Terminus to debug a piece of code that wasn't working. In chat mode, it found the bug and fixed it. Quick and simple. In reasoner mode, it explained why the bug happened, how to fix it, and how to prevent similar bugs in the future. Both answers were correct. But reason mode taught me something new. That's the power of having both modes available. Use the right tool for the job. Terminus uses an MIT license that's very permissive. You can use it for commercial projects without issues, but always check the terms of service if you're using hosted APIs. Different providers have different rules. The community response has been mostly positive. People love the language fixes. The agent improvements get good reviews, but some users report occasional issues. Hallucinations aren't completely gone. Some complex prompts still cause problems. My advice, test thoroughly with your specific use cases. Don't assume it works perfectly for everything. For coding tasks, Terminus is solid. The code quality is noticeably better. Error handling improved. Comments are more useful. For research tasks, the search agent upgrades are real. Better information retrieval. More relevant results. For general chat, the language fixes make conversation smoother. No more weird character mixing issues. But remember, this is still an AI model. Verify important outputs. Don't trust blindly. If you want to try Terminus yourself, start with the API. It's the easiest way to test the capabilities. The hog inface model page has all the technical details. The model card explains the improvements. Inference examples show you how to use it. For beginners, I recommend starting with simple prompts. Test both chat and reasoner modes. See which one fits your workflow. Advanced users can dive into local installation, but expect to need powerful hardware and patience for downloads. Let me show you one more demo before we wrap up. I'm going to ask Terminus to write a complete web scraper in Python. Watch how it handles this complex request. It needs to understand web scraping, Python libraries, error handling, and best practices. Look at this output. Clean imports, proper error handling, good documentation, rate limiting to be respectful. This isn't just working code. is production ready code with best practices built in. The old version would give you working code, but Terminus gives you professional code. And the explanations are clearer, too. It doesn't just tell you what the code does. It explains why it made specific choices for anyone learning to code or building AI applications. These improvements are huge. Better code quality means fewer bugs. Better explanations mean faster learning. So, should you upgrade to Terminus if you're already using DeepSeek? Absolutely. The fixes alone justify the switch. If you're new to Deepseek, Terminus is a great entry point. It shows the model at its best. For coding tasks, it's competitive with other top models. For research, the search agent improvements are noticeable. The two mode system gives you flexibility, quick answers when you need them, deep thinking when you want it. Just remember the hardware requirements for local runs. A PI access is much more practical for most users. For developers and researchers, the improvements are worth the upgrade. For casual users, the cleaner responses alone make it worthwhile. If you found this useful, drop a comment with prompts you want me to test. Julian reads every comment, and your suggestions drive future videos. Now, if you want to take your AI game to the next level, I've created the free AI money lab where I share exactly how to profit from AI tools like Deep Seek Terminus. Inside, you'll get 50 plus free AI tools and 200 plus chat GBT SEO prompts. You'll learn how to make money with AI agents, get access to 1,000 plus free automation workflows, and see how one member made 10,000 plus with chat GPT. Plus, you get a full blueprint to generate thousands of leads free with AI. What you'll also get is access to our free AI community, a complete AI course, and proven AI case studies. The link is in the description. It's completely free and packed with value. The future of AI is open- source models competing with closed source giants. Deepseek Terminus shows that future is closer than most people think. Thanks for watching and I'll see you in the next one.