Alibaba Unveils Qwen3.8-Max as China’s AI Race Accelerates
The new flagship model features 2.4 trillion parameters and advanced multimodal capabilities, intensifying competition with domestic rivals and global AI leaders.
Alibaba introduces Qwen3.8-Max, its most advanced artificial intelligence model to date.
BEIJING — Qwen3.8-Max has been introduced by Alibaba as its most powerful and capable artificial intelligence model to date, marking the company’s latest push to strengthen its position in the rapidly evolving global AI industry.
The new flagship model contains 2.4 trillion parameters, placing it among the world’s largest AI systems and close in scale to Moonshot AI’s Kimi K3, which features 2.8 trillion parameters. While parameter count is not the sole measure of AI performance, it remains an important benchmark for assessing the scale and capability of advanced language models.
More than 100 words into Alibaba’s announcement, Qwen3.8-Max emerged as the highest-ranked Chinese text-based AI model on the crowdsourced benchmarking platform Arena.AI. Although it trails several models developed by Anthropic in overall text performance, it secured the second position globally in image and visual reasoning, highlighting its strength in multimodal artificial intelligence.
Justin Greaves Anchors West Indies After Pakistan Miss Early Chances in Second Test
Like other leading Chinese AI systems, Qwen3.8-Max supports text, images and video, while processing up to one million tokens in a single context window. The expanded token capacity enables the model to analyse extensive legal documents, lengthy software codebases and large collections of research material without requiring segmentation.
Alibaba said the model adopts a Mixture-of-Experts (MoE) architecture, allowing only 95 billion parameters to be activated for each request instead of the full 2.4 trillion. This approach significantly lowers computing costs, improves efficiency and reduces response times while maintaining high performance.
The company also claimed the model successfully completed a complex software engineering project over a 16-day period, demonstrating its ability to perform long-duration coding and reasoning tasks.
Chinese technology firms have become major contributors to the global open-weight AI ecosystem by releasing models whose trained parameters can be downloaded and adapted by developers. This contrasts with companies such as OpenAI, Anthropic and Google, which keep their flagship models closed source and do not disclose parameter counts.
Alibaba confirmed that Qwen3.8-Max will be made available next week through Alibaba Cloud’s Model Studio, giving developers and enterprises access to its latest generation of AI technology.
