Qwen 3.8 Next First Test | Coding, Game Dev, Frontend | 2.4T Parameter Open Weight model by Alibaba

· Source: Venelin Valkov · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Software Development & Engineering · Depth: Intermediate, medium

Summary

Alibaba's Qwen 3.8 Max Preview, a large language model with approximately 2.4 trillion parameters, was tested for its coding capabilities in game development and frontend design. Despite Alibaba's claim of its compatibility with leading frontier AI models, second only to Fable 5, the model demonstrated inconsistent performance. Initial tests, such as generating a Snake game, resulted in the model getting stuck in a thinking loop for over seven minutes without producing code. While it successfully generated HTML for an embodied gravity simulation and a boating game, the output speed was slow (around 20-25 tokens/second), and the quality, physics, and design were deemed inferior to models like Fable 5, Kimik A3, and GM 5.2. A final test for an ML/AI engineer website also yielded a "pretty bad" design compared to competitors, leading to an overall unimpressed assessment by the reviewer.

Key takeaway

For AI Engineers evaluating new open-weight models for coding and frontend development, Qwen 3.8 Max Preview, despite its 2.4 trillion parameters, currently underperforms leading frontier models. You should exercise caution and conduct extensive testing, especially for complex tasks involving game logic or intricate UI design, as it exhibited issues like getting stuck in thinking loops and producing suboptimal code. Consider alternatives for production-grade coding applications.

Key insights

Alibaba's 2.4T parameter Qwen 3.8 Max Preview model shows inconsistent coding performance, often struggling with complex tasks and exhibiting slow "thinking."

Principles

Method

The author tested Qwen 3.8 Max Preview via its chat interface using heavy prompts for game development (Snake, boating) and frontend design (gravity simulation, engineer website), observing output quality, speed, and internal "thinking" processes.

In practice

Topics

Best for: AI Engineer, Machine Learning Engineer, AI Scientist

Related on AIssential

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Venelin Valkov.