Alibaba Unveils Qwen 3.8: A New 2.4 Trillion-Parameter AI Giant
The AI race between global tech superpowers just got more intense. Alibaba has previewed its latest AI model, Qwen 3.8, claiming it is one of the most powerful models available today. The company stated that Qwen 3.8 trails only Anthropic's Claude Fable 5 in performance. The announcement was made at the World AI Conference (WAIC) in Shanghai, signaling how quickly Chinese AI labs are closing the gap with their American counterparts.
Unpacking Qwen 3.8's Capabilities
According to Alibaba, Qwen 3.8 brings significant advancements:
The model carries a massive 2.4 trillion parameters, a measure of an AI model's complexity.
It significantly outperforms its predecessor, Qwen 3.7 Max, across coding, reasoning, and complex tasks.
Targeted workloads include full-stack software development, data analysis, and office workflows.
A preview of the model is already available on Alibaba's Token Plan subscription service and its Qoder and QoderWork platforms. Furthermore, Alibaba revealed that Qwen 3.8 will soon be released as an open-weight model, meaning developers will gain access to the model's learned parameters while its training data and source code remain proprietary.
A Rivalry Taking Shape in the AI Space
Alibaba's launch appears partly timed to counter the momentum of Moonshot AI, a Beijing-based lab that unveiled its own massive model—Kimi K3—just days earlier. Kimi K3 is larger, with 2.8 trillion parameters, making it currently the world's biggest open-source AI model.
Moonshot stated that K3 delivers frontier-level performance across a wide range of benchmarks and outperformed both Anthropic's Claude Fable 5 and OpenAI's GPT-5.6 Sol on tests measuring an AI's ability to build software projects from scratch and handle long-horizon software engineering tasks. Those results, however, are based on Moonshot's own self-reported data. In other words, they are not verified by independent benchmark organizations.
The same goes for Alibaba's claim that Qwen 3.8 is stronger than K3, which has also not yet been backed by independent evaluations.
The Multi-Trillion Parameter Era
Both launches point to a broader trend: Chinese AI labs have now entered what is being described as the multi-trillion parameter era, and are benchmarking themselves directly against the world's most advanced AI companies — Anthropic, OpenAI, and Google.
The launches also carry a broader political signal. Chinese President Xi Jinping made his first in-person appearance at WAIC this year and said China was committed to "encouraging open source, openness, collaboration and sharing," adding that such an approach could help ensure "all sectors and businesses can benefit from AI".