[00:03] Did you hear what Alibaba just dropped? Most people are scrolling right past it. That might be a big mistake because this model is built differently from the ones before it, and the real story is not what the headlines say. Let me show you. [00:16] Today I'm breaking down Qwen 1.5 Max, what it is, how it works, and who can use it right now. And stick around to the end because there's one thing you should check before you trust the big numbers. First, the basics. On November [00:30] numbers. First, the basics. On November 16th, 2026, Alibaba launched Qwen 1.5 Max. It's the top model in the new Qwen 1.5 family. In its announcement on X, the Qwen team called it your new AI teammate for work, creativity, and [00:44] discovery. This didn't come out of nowhere. A limited preview called Qwen 1.5 Max preview opened on November 11th. Now the full model is live. So what is it? It's a multimodal model. That just means it can take in text, images, and [00:59] video. What you get back is text. Alibaba built it for coding, office work, front-end design, and complex agent tasks. That's straight from the model page. An agent task is when an AI works through many steps on its own to [01:13] finish a job. That's where this model is aimed. Now here's where it gets interesting. Alibaba says it has 2.4 trillion parameters. A parameter is one of the numbers the model learned in [01:25] more capacity. design. That means only part of the network runs for each request. So you get a huge brain without waking up the whole thing [01:39] every single time. It also supports a context window of 1 million tokens. A token is a small chunk of text, like a word or part of a word. So a million tokens means the model can hold a very large amount of text in its head at [01:53] once. That's a lot of room for long documents, long chats, or big projects. On the model page, max input is listed at 991,000 at 991,000 tokens, and the max output at 393,000. [02:07] That's the foundation. Next up, what you can expect from it. Let's start with efficiency. Alibaba compared it with its last flagship, Qwen 3.8 Max. Their report says Qwen 4 Max uses up to about half the tokens to do the same kind of [02:22] work. Fewer tokens means less compute for every task. That matters a lot for anything long-running, like agents that go through many steps. Next, quality. Alibaba built its own test called Work 360. It checks tool use and long [02:38] professional tasks. On that test, Qwen 4 Max scored 52.69. Max scored 52.69. Qwen 3.8 Max scored 47.74. So that's a clear step up on Alibaba's own test. Now, coding. Alibaba says Qwen [02:54] 4 Max scored near the top of coding benchmarks and beat rival models like Kim K 3.1 Pro. On Code Arena, a blind test for front-end coding, it was reported to rank sixth in the world. Quick pause before I show you the next [03:08] part. We've got a 30-day roadmap inside the AI Profit boardroom that helps you learn new AI tools like this one, one step at a time. There are four coaching calls every week. So if you get stuck on a new model or a new setup, you can [03:22] bring your questions and get unstuck. The roadmap is built to help you turn tools like this into a real skill you can use in your own workflow. No guessing what to learn next. Link in the description or go to [03:34] aiprofitboardroom.com. Those are big claims, and I'll come back to one catch in a minute. First, let's look at how it works. If you're a developer, you call Qwen 4 Max through an API. The model name is Qwen 4 [03:47] Max. It's compatible with both the OpenAI and the Anthropic API styles. So if your tools already speak one of those, plugging it in is simple. It also has a thinking mode. When you turn it on, the model shows its reasoning before [04:01] the final answer. On the model page, the max reasoning length is listed at 262,000 tokens. Then, there are the built-in features. Function calling lets the model connect to outside tools. [04:14] Structured outputs make it return clean JSON in the format you ask for. Web search lets it pull in live information. Context cache stores the shared parts of long prompts, so repeat requests run faster. Batches let you send lots of [04:28] requests at once and get the results later. Prefix completion makes the model continue strictly from text you give it, and fine tuning lets you train it on your own sample data for a specific task. There are also built-in tools on [04:42] the responses API. That includes a code interpreter, web search, and a web pages. So how do you actually get started? Here's the simple path from the token plan page. Once you have a Qwen Cloud [04:56] account and a plan in place, you get your base URL and API key from the API keys page. Then you open a supported tool, like Cline, Cursor, or Claude Code, and enter that base URL and key in its settings. That's it. If you use Qwen [05:11] Code, it's even simpler. Open settings, find the provider section, choose the Qwen Cloud coding plan, and sign in with your Qwen Cloud account. And if you're calling the API directly, the model name to use is Qwen 4 Max, and you add enable [05:26] thinking to turn on the thinking mode. One more number from the model page. The rate limit is listed at 5 million tokens per minute. That's how much text you can matters if you're running a lot of agents at once. Now, a few more things [05:41] worth knowing. Alibaba also has a token plan on Qwen Cloud. It's a plan that gives you access to multiple models in one place. The higher tiers let you run more agents at the same time. The light plan runs one to two agents, and the pro [05:56] plan runs six to eight. Reports say Qwen 4 Max launched with token plans available too. The plan works with tools that support the OpenAI and Anthropic styles. That includes Cuen code, Klein, Claude code, cursor, codex, and open [06:11] code. So is it open source? Not this one. Cuen 4 Max is a proprietary model. Alibaba has said smaller Cuen 4 models will be open source later, but there's no release date yet. So who is this for? First, developers. If you build apps or [06:27] agents, you can call the API today. Second, people who live inside coding tools. Because it plugs into tools like cursor, Klein, and Claude code, you can try it inside the setup you already use. Third, anyone working with long text [06:43] because of that million token window. And fourth, anyone who just wants to see it for themselves. The model page has a try AI chat where you can test Cuen 4 Max directly. One thing to remember, it takes in text, images, and video, but it [06:57] only gives text back. So it can look at an image or a video and reason about it, itself. Let's wrap up. Cuen 4 Max is Alibaba's [07:09] new flagship. It has 2.4 trillion parameters, a sparse design, native vision, and a million token window. Alibaba says it uses far fewer tokens and scores higher than its last model. Now, the catch I promised you. Those [07:25] numbers are Alibaba's own. Reports say independent rankings are less generous, and I couldn't find anyone outside Alibaba who has rerun work 360. So don't take the big numbers as the final word. Test it on your own [07:37] tasks first. That's how you learn what it's actually good at. Right now, owners haven't figured this out yet, and that window won't stay open forever. The people who learn these tools early get ahead on understanding them. Inside the [07:52] AI Profit boardroom, you get daily tutorials, so there's always something calls every week where you can bring your questions. You get a prompt library with prompts to start from, and you get the member map, a guide to what's inside [08:07] and where to begin. Link in the description or go to AIprofitboardroom.com. If you want the full process, the SOPs If you want the full process, the SOPs and over 100 AI use cases like this one, [08:19] join the AI Success Lab. It's our free community. Links in the comments and description. You'll get all the video notes from this episode plus access to 95,200 members. Joining is free. The AI Success [08:32] Lab is a place to learn AI alongside other people working through the same tools. The video notes from this episode give you a written version of what we covered so you can come back to it anytime. Join today and I'll see you in [08:44] anytime. Join today and I'll see you in the next video.