Alibaba Keeps Qwen4-Max Closed and Sells Access at $1.50 per Million Input Tokens

Alibaba's 2.4-trillion-parameter Qwen4-Max is offered only as a hosted model, at $1.50 per million input tokens. The company's own benchmark chart shows it leading on some coding and legal tests and trailing Claude Opus 5.5 on others.
Abstract illustration of a dark lattice of nodes with a few glowing amber nodes linked to one bright blue source, suggesting sparse expert routing
Alibaba Keeps Qwen4-Max Closed and Sells Access at $1.50 per Million Input Tokens
At a Glance
  • Price: $1.50 per million input tokens, $4.50 output (Singapore)
  • Weights: Closed; smaller Qwen4 models to get open weights later
  • Size: 2.4 trillion backbone parameters, 71 billion active per token
  • Cost claim: Alibaba says Work360 task cost fell to $1.67 from $4.43
  • Rankings: No. 14 on Arena’s Agent Arena, top Chinese model

Alibaba’s most capable AI model is one developers can’t download. Qwen4-Max, the 2.4-trillion-parameter flagship that the company’s Qwen team released late Sunday U.S. time, is offered only as a hosted model, through Alibaba’s QwenCloud API and apps including Qwen Studio, QwenWork and Qoder. Smaller Qwen4 models will get open weights later, the team said, without giving a date.

That is a reversal from the last generation. The core of Qwen3.8-Max shipped with open weights in August as Qwen3.8-2.4T-A95B, a text-only release and the first Max-class Qwen to be open. Qwen is already the most downloaded open-weight model family in the world, according to Bloomberg, as cited by TNW. Alibaba is selling the closed model on price. QwenCloud’s pricing page lists $1.50 per million input tokens and $4.50 per million output tokens in its Singapore region, with cache reads at 7.5 cents per million. Beijing and global deployments run slightly lower, at $1.413 and $4.24, and thinking and non-thinking modes cost the same. OpenAI lists GPT-6.1 Sol at $2 and $10 and its mid-tier GPT-6.1 Terra at $1 and $5, as WebProNews reported when OpenAI priced GPT-6.1 Terra last month.

Credit: Qwen (@Alibaba_Qwen). View on X.

The Qwen team’s launch post says the model also does more with fewer tokens than Qwen3.8-Max. On Work360, Alibaba’s in-house benchmark of industry tasks across 721 specialized fields, token use fell about 35% at High reasoning effort and 24% at Xhigh, and the average cost per task dropped from $4.43 to $1.67. On Terminal-Bench 4.0, token use fell by 48% to 49%.

Under the hood, Qwen4-Max is a mixture-of-experts model that activates 71 billion of its 2.4 trillion backbone parameters per token, 25% fewer active parameters than Qwen3.8-Max at the same scale, Alibaba Cloud said. It reads images and video natively and takes up to 1 million tokens of context. Thinking plus the final answer is capped at 262,144 tokens, and fine-tuning and batch inference aren’t supported, the developer docs say.

Strong on some tests, behind on others

Alibaba’s own comparison chart is mixed. Qwen4-Max scored 75.7 on the DeepSWE coding benchmark, a hair ahead of Claude Opus 5.5 at 74.2 and Kimi K3.1 at 75.5, and 21.3 on the HLAB-Vals legal agent test, where no rival in the chart topped 11.7. It trailed on Terminal-Bench 4.0 with 48.5, against 59.6 for Opus 5.5 and 67.2 for Kimi K3.1, and on ProgramBench, a black-box software replication test, it posted 31.5 to Opus 5.5’s 65.0.

Grouped bar chart comparing Alibaba-reported scores for Qwen4-Max, Claude Opus 5.5, GPT-6.1 Sol and Kimi K3.1 on DeepSWE, Terminal-Bench 4.0, ProgramBench, JobBench, HLAB-Vals and Humanity's Last Exam

Outside leaderboards rank it as the strongest Chinese model, behind several U.S. rivals. Arena said Qwen4-Max debuted at No. 14 overall on its Agent Arena and is the top Chinese model there, and ranked Alibaba the No. 5 lab overall, behind Anthropic, Google, OpenAI and Meta. On Code Arena’s WebDev board it landed at No. 6 with a preliminary 1,750 points. Vals AI ranked it 11th on the Vals Index, also the highest of any Chinese model, with notable gains over Qwen3.8-Max on coding and finance tasks.

Long agent runs and a filtered cyber diet

Alibaba is pitching the model for agents that work around the clock. In one demonstration, Qwen4-Max ran for 48 hours on Deep Principle’s Mira lab platform, built a candidate space of 2.7 million molecules and directed an automated wet lab through four design-and-test loops before producing an experimentally validated chip coolant molecule, the Qwen team wrote. In a test in automotive production at SANF and Pacific Union, the manual effort to program a single workpiece fell from about 475 minutes to 24. The team also open-sourced a Live Delivery plugin that has an agent hand over a usable first version early and revise it as users step in.

On security, Alibaba said it filtered cyber training data by task intent, keeping material for vulnerability detection and remediation and removing data meant to strengthen attacks. A review by its internal security experts found that “no capabilities posing large-scale cybersecurity risks have been observed,” the post says. Moonshot AI took a different route last month, removing cyber-related data from training ahead of Kimi K3.1 Pro’s open-weight release. Alibaba also reported that Qwen4-Max’s pass rate for code that is both functional and secure on Last Secure Code Bench rose to 31.8%, from 22.0% for Qwen3.8-Max.

Developers can already reach the model outside Alibaba’s cloud. Qwen said it is live on OpenRouter and Vercel’s AI Gateway, and QwenCloud serves it from Singapore, Beijing, Frankfurt, Hong Kong, Tokyo and Virginia. Alibaba hasn’t said when the open-weight Qwen4 models will arrive.

Subscribe for Updates

GenAIPro Newsletter

News, updates and trends in generative AI for the Tech and AI leaders and architects.

By signing up for our newsletter you agree to receive content related to ientry.com / webpronews.com and our affiliate partners. For additional information refer to our terms of service.

Notice an error?

Help us improve our content by reporting any issues you find.

Get the WebProNews newsletter delivered to your inbox

Get the free daily newsletter read by decision makers

Subscribe
Advertise with Us

Ready to get started?

Get our media kit

Advertise with Us