Alibaba Launches Its Most Powerful Qwen Model Yet
Alibaba has unveiled Qwen3.8-Max, its largest artificial intelligence model yet, as Chinese developers intensify efforts to challenge leading US systems with high-performance open-weight models that are cheaper to run.
Topics
News
- Temasek, Seraphim Back Pixxel in $100 Million Round
- UN Rights Chief Warns Advanced AI Could Threaten Humanity
- Tata-owned JLR to Cut About 4,000 Bobs Amid Tariff, China Pressure
- Anthropic Walks Away From $6 Billion Decart Deal
- Hyundai India Targets 20% Women in Executive Workforce by 2030
- Security Breach Drains $320 Million in Bitcoin From Liquid Network
[Image: Nomita Samaiyar/MITSMR India]
Alibaba has launched Qwen3.8-Max, its largest and most capable artificial intelligence model, as Chinese developers race to improve performance while containing operating costs.
The model has 2.4 trillion parameters, slightly fewer than the 2.8 trillion in Moonshot AI’s Kimi K3. Parameters are numerical values learned during training, although a higher count does not necessarily produce better results.
Qwen3.8-Max accepts text, images and video and supports a context window of up to one million tokens. That allows it to examine large code repositories, lengthy legal files or hundreds of pages of documents in a single request.
The model is available to global developers through Alibaba Cloud’s Model Studio APIs and through QwenWork, its workplace AI-agent platform.
The company plans to release the model weights next week, allowing developers to download and adapt the underlying system. It is also available through QwenWork, Alibaba’s workplace AI-agent platform.
Qwen3.8-Max ranked fifth on Arena’s text-model leaderboard, making it the highest-ranked Chinese model, and second on its vision leaderboard. The rankings are based on users comparing model outputs, remain preliminary and may change as more votes are recorded.
Despite its total size, the model activates only 95 billion parameters for each request. Its mixture-of-experts architecture assigns work to specialized parts of the model rather than running the entire system, reducing computing costs and response times.
Alibaba said Qwen3.8-Max autonomously completed a software-engineering project over 16 days, repeatedly generating code, running tests and analyzing results. The claim comes from an internal evaluation and has not been independently verified.
The launch reflects a broader push by Chinese AI developers to win users with open-weight models. Once released, the weights will allow developers to run and modify Qwen3.8-Max, unlike the proprietary flagship systems offered by OpenAI, Anthropic and Google.
Alibaba’s Hong Kong-listed shares rose 7% after the announcement.
