Back

Alibaba Releases Qwen4-Max Through QwenCloud

Alibaba released Qwen4-Max on November 16, 2026, with access through QwenCloud and its QwenWork and QwenStudio platforms, according to the company. Alibaba said the new release ranked above all earlier models in its Qwen family for capability.

On QwenCloud, the company said, access was available through both the API and Token Plan. Alibaba put the model’s size at 2.4 trillion parameters and said its features included vision capabilities and a context window supporting as many as one million tokens.

The general release came after a limited preview that started on November 11, according to TechNode.

For its comparison with Qwen3.8-Max, Alibaba published results from Work360, its proprietary benchmark. The company said the assessment covered difficult industry tasks across 721 specialized fields and that Qwen4-Max performed better than Qwen3.8-Max. It reported that token consumption was 35% lower when using high-effort reasoning mode, while the average cost for each task fell by about 60%.

Alibaba also outlined changes to the material and checks used during training. The company said it kept cybersecurity content designed to support the detection and repair of software vulnerabilities, but excluded material directed at offensive attacks. Its software-engineering training also included checks for code security, according to Alibaba.

Separately, Reuters reported a company-supplied secure-code-generation benchmark result of 14.7% for Qwen4-Max, against 8% for its predecessor. That assessment covered functions, files and software repositories. Reuters said the findings had not been independently verified by the news agency.

Alibaba said its safety work also addressed chemical, biological, radiological and nuclear risks. The company stated that it added safety datasets and conducted post-training work intended to strengthen protections in those areas.

* This article is based on publicly available information at the time of writing.