Alibaba Launches Qwen4-Max to Challenge U.S. AI Leaders
Alibaba Group Holding launched the Qwen4-Max flagship AI model featuring a 2.4 trillion parameter architecture to compete with OpenAI and Anthropic.
Alibaba Group Holding launched Qwen4-Max on November 16, 2026, a flagship artificial intelligence model designed to challenge U.S. leaders OpenAI and Anthropic. The model utilizes a sparse Mixture-of-Experts architecture with 2.4 trillion total parameters, activating 71 billion parameters per token. Alibaba claims the model outperforms Anthropic’s Claude Fable 5.1 and Moonshot AI’s Kimi K3.1 Pro in coding benchmarks.
Qwen4-Max features a one-million-token context window and native vision capabilities, targeting professional workflows in finance, law, healthcare, and automotive manufacturing. Compared to its predecessor, Qwen3.8-Max, the new model reportedly reduces average task costs by 60% and uses 35% fewer tokens in high-effort reasoning mode. While Qwen4-Max is not open-weight, the company plans to release smaller open-weight versions of the series to developers soon.
Under CEO Eddie Wu, the company is building a comprehensive AI ecosystem that includes a target of 20 gigawatts of data center capacity by 2032 and the production of domestic AI chips via its T-Head unit. To mitigate cybersecurity risks, Alibaba filtered training data to remove offensive cyberattack capabilities while retaining defensive tools for vulnerability detection and implementing safeguards against chemical, biological, radiological, and nuclear risks. The model is available via QwenCloud, QwenWork, and QwenStudio.