Article complete

Alibaba Launches Qwen4-Max AI Model With Focus on Security

The 2.4-trillion-parameter model includes a 1-million-token context window, native vision capabilities and security improvements across coding and other tasks.

alibaba

Key takeaways

  • Alibaba has launched the Qwen4-Max AI model, featuring 2.4 trillion parameters and a 1-million-token context window.
  • The model is designed for a wide range of applications, including 3D modelling, finance, and healthcare.
  • Qwen4-Max outperforms its predecessor, Qwen3.8-Max, in nearly every evaluated capability, with reduced token usage.

Takeaways drafted with AI from the article text.

Alibaba has introduced Qwen4-Max, its latest flagship artificial intelligence (AI) model. Built on a new architecture with 2.4 trillion parameters and a 1-million-token context window, the model is designed for complex professional and scientific tasks.

“It can take on a wide range of creative and professional work,” Alibaba said, citing applications in 3D modelling, game development, precision manufacturing, finance, law, healthcare, and scientific discovery.

Qwen4-Max succeeds Qwen3.8-Max, which Alibaba launched in August. The company said the new model outperforms its predecessor across nearly every evaluated capability. In its Work360 evaluation, token usage fell by approximately 24% and 35% at Xhigh and High effort levels, respectively, the company claimed.

Alibaba also reduced API prices, particularly for cached inputs. The average cost per task in its Work360 evaluation fell from $4.43 to $1.67—a 62% reduction. The earlier Qwen3.8-Max launch marked another step in Alibaba's expansion of its AI model portfolio.

The company also highlighted improvements in security testing. On Last Secure Code Bench, Qwen4-Max's function-level pass rate for both functionality and security increased from 22% for Qwen3.8-Max to 31.8%. The share of code meeting security requirements at the function, file, and repository levels rose from 8% to 14.7%.

Alibaba said it had strengthened safeguards against cybersecurity misuse by retaining training data for defensive tasks, such as vulnerability detection and remediation, while removing data intended to improve offensive attack capabilities. Internal testing found no capabilities posing large-scale cybersecurity risks.

For chemical, biological, radiological, and nuclear risks, Alibaba introduced dedicated safety data during post-training. The company said internal evaluations found that the model identified more potentially harmful requests and provided safe responses, without significantly increasing users' ability to carry out harmful tasks.

The security focus comes after scrutiny of earlier Qwen models. A Cohere security evaluation of an Alibaba Qwen model raised concerns about the model's behaviour during testing.

Alibaba's latest release comes amid growing competition among Chinese technology companies to improve AI model performance while reducing operating costs. Several Chinese developers have also gained prominence in the global open-weight AI market.

Qwen4-Max is available through Alibaba's QwenCloud platform, QwenWork, Qoder, and QwenStudio. Developers can access it through a token plan or API.

Published

Corrections: AIM corrects errors of fact and notes the change on the story. Spotted one? Tell the editors. Read our editorial standards.

More in this story

AI in Healthcare · 5 stories
  1. ServiceNow Partners with Anthropic to Deploy Claude For 29,000 Employees
  2. YouTube Expands AI Deepfake Detection Tool to Politicians, Officials & Journalists
  3. AI Is Making Cheaper MRI Machines Look Premium
  4. Meta Expands AI Age-Estimation Tech to Protect Teens on Instagram
  5. You are here: Alibaba Launches Qwen4-Max AI Model With Focus on Security
About the author

Senior journalist

Mukundan covers Indian IT, enterprise AI and the AI startup ecosystem for AIM. Reach out to him at or Signal at mukundan.42.

All stories
Belamy · Daily

The day’s AI news, before your first meeting.