01What Is China's Qwen 3.8 Max AI?

China's Qwen 3.8 Max AI is a frontier-level artificial intelligence model released by China. It has 2.4 trillion parameters and is designed to handle complex tasks like coding, game creation, and multimodal understanding. This model is open-weight, meaning you can access it freely for now.

In this episode, Julian Goldie explains that Qwen 3.8 Max performed very well on benchmarks. It even beat several leading models in coding and vision tests. The model also coded autonomously for 10 to 16 days, improving its own code on a GitHub repo.

02Main Features of Qwen 3.8 Max AI

Qwen 3.8 Max comes with several powerful features. First, it is a multimodal model, meaning it can process text, images, and other formats. This makes it useful for a wide range of applications.

  • Autonomous coding: The model can code for days without human help. In tests, it ran for 16 days, self-evolving on GitHub.
  • High benchmark performance: It ranks in the top three on many tests, including front-end code and text arena. It beats models like Opus 5 and Gemini 3.1 Pro.
  • Free access: You can use it at chat.quen.ai without paying. API access is also available on Quen Cloud.
  • Open weight release: A local 27B model is coming next week, so you can run it on your own computer.

03How to Set Up Qwen 3.8 Max AI

Setting up Qwen 3.8 Max is simple. Julian shows two main ways to get started. Follow these steps:

  1. Go to chat.quen.ai in your browser. You can use the model for free right away.
  2. If you need API access, sign up on Quen Cloud. This lets you call the model from your own apps.
  3. To use it inside advanced tools, get the Agentic Operating System from aiprofitbomb.com. The link is in the episode description.
  4. Download the full agent system zip file. It includes setup files for Qwen 3.8.
  5. Watch the included one-hour masterclass course to learn how to use Qwen 3.8 Max directly.

04Using Qwen 3.8 Max to Build and Test

Julian demonstrates how Qwen 3.8 Max can build games, websites, apps, and landing pages. He tested it by creating 3D games like a Skyrim-style open world. The model handles physics, coding logic, and UI design well.

You can also plug it into Claude Code inside the Agentic Operating System. This lets you use Qwen 3.8 alongside other tools like Kimi K3. Julian ran the "Goldie Bench" test, where Qwen 3.8 Max performed very well.

The model feels more open and creative than others like Fable 5, which builds more linear projects.

To see what others created, check the autonomous GitHub repo. Julian notes that the quality of 3D worlds and game interfaces is impressive for a model that just launched.

05Test Results and Benchmark Comparisons

Julian shares multiple benchmark results from the episode. On the front-end code leaderboard from Arena AI, Qwen 3.8 Max ranks fourth, above Fable 5. On the text arena, it comes in fifth, beating Gemini 3.1 Pro and GPT 5.6 Soul.

On the vision arena, it ranks second, only behind Fable 5.

Overall, Qwen 3.8 Max is consistently in the top three across tests. The only model that outperforms it in most cases is Fable 5. Julian emphasizes that this is impressive for an open-source Chinese model.

The model also coded autonomously for 16 days, looping and improving its own GitHub repository.

06Tips, Limits, and Mistakes to Avoid

Julian provides several tips for using Qwen 3.8 Max effectively. First, test it on games or UI projects to see its creative range. Use the free chat interface at chat.quen.ai before committing to API access.

One limit is that the 27B local model is not yet out. The current model may require cloud resources. Julian warns against comparing it only to older models like Opus 4.8.

Instead, compare it to frontier models like Fable 5. Also, avoid using it in linear tests only; its strength is in open-ended tasks.

A common mistake is ignoring the autonomous coding feature. You can let it run for days on a repo. Make sure to use the GitHub integration to see its progress.

07Who Should Use Qwen 3.8 Max AI and Why It Matters

Qwen 3.8 Max is for developers, AI enthusiasts, and builders who want a frontier model without paying high fees. Julian points out that it is open-weight, so you can use it freely. The ability to code autonomously for days makes it ideal for long-running projects.

It also matters because it shows that open-source Chinese AI can compete with top Western models. The soon-to-be-released local 27B model will make it accessible for more people. Julian recommends joining the aiprofit boardroom community to get the full agent system and learn advanced workflows.