Alibaba Unveils Its Most Advanced AI Model Yet, Narrowing the Gap With Moonshot’s Model in Scale
China’s Alibaba on Monday introduced its largest and most advanced artificial intelligence model to date, Qwen3.8-Max, saying the system is comparable in scale to a model unveiled by domestic rival Moonshot AI last month.
Chinese technology companies, a major force in the global open-weight artificial intelligence market, are engaged in an intense race to develop increasingly powerful AI systems while keeping the cost of operating them under control.
Alibaba’s Qwen3.8-Max has 2.4 trillion parameters — numerical values that AI models learn from data and use to identify patterns, generate responses and perform tasks. By comparison, rival Moonshot AI’s Kimi K3 has 2.8 trillion parameters.
While a higher parameter count does not necessarily mean a better-performing model, it is widely viewed as an indicator of the computing power and data involved in developing advanced AI systems.
Chinese technology firms often disclose parameter counts to encourage adoption among developers. Many of their AI models use an open-weight approach, allowing developers to download the underlying learned parameters and use or modify the systems.
By comparison, OpenAI, Anthropic and Google do not disclose parameter counts for their closed-source AI models.
Qwen3.8-Max was launched on Arena.AI, a crowdsourced platform for comparing AI models, where it quickly became the highest-ranked Chinese text model. However, it remained behind Anthropic’s Claude Fable 5 and three Opus variants.
On Arena.AI’s leaderboard for models capable of analysing images and other visual content, Qwen3.8-Max ranked second globally, trailing only a variant of Claude Fable 5.
Both Qwen3.8-Max and Kimi K3 are capable of processing text, images and video, with each able to handle up to 1 million tokens in a single context.
Tokens are units of data that can represent parts of words or short words. A larger context window allows AI systems to process substantial amounts of information at once, including lengthy legal documents, large software codebases and hundreds of pages of text.
Alibaba said Qwen3.8-Max uses a “mixture-of-experts” architecture, which assigns tasks to specialised components rather than activating the entire model for every request. The company said only about 95 billion parameters are activated at any given time, helping reduce operating costs and response times.
Alibaba also said the model completed a software-engineering project in 16 days.
Qwen3.8-Max is scheduled to become available next week through Alibaba Cloud’s Model Studio platform.
