Alibaba makes Qwen4-Max generally available at $1.50 per million input tokens
Alibaba released its Qwen4-Max flagship model to all API customers on November 16, five days after a limited preview. The closed-weight model is priced 25% below its predecessor and trails US frontier models on Alibaba's own coding chart.
Cortex, covering AI models, labs and agents
, 3 min read
$1.50per million input tokens; $4.50 per million output
What to know
- Qwen4-Max became generally available on November 16 through Alibaba's QwenCloud API and Model Studio, after a preview that opened November 11.
- It is a closed-weight mixture-of-experts model with 2.4 trillion total parameters, 71 billion active per token, and a 1 million token context window.
- The international list price is $1.50 per million input tokens and $4.50 per million output tokens, down from $2.00 and $6.00 for Qwen3.8-Max.
Source: DataCamp, November 16, 2026, relaying Alibaba's published chart.
Alibaba made Qwen4-Max, the flagship of its Qwen4 model family, generally available on Monday, November 16. The model had been open since November 11 only as a preview through Alibaba's subscription Token Plan, according to DataCamp and Data Studios. It is now sold per token through the QwenCloud API and Alibaba Cloud Model Studio under the model ID qwen4-max, and it runs in Alibaba's Qwen Studio, QwenWork and Qoder apps.
Alibaba's own announcement page could not be opened for this story. The figures below come from DataCamp, Digital Trends and Data Studios, which all reported on the launch on November 16.
What Alibaba released
Qwen4-Max is a mixture-of-experts model with 2.4 trillion total parameters, of which 71 billion are active for each token, according to Digital Trends and DataCamp. It accepts text, images and video and replies in text. The context window is 1 million tokens, and the model "thinks" before answering by default, with low, high and extra-high effort settings, Data Studios reported. (For what that thinking step does, see our reasoning models explainer.)
The weights are closed. Alibaba says smaller Qwen4 models will be released with open weights later but has given no dates, according to all three reports. That breaks with the pattern of many earlier Qwen releases; our open-weight versus closed explainer covers what the difference means for developers.
Price
In Alibaba's Singapore (international) region, Qwen4-Max costs $1.50 per million input tokens and $4.50 per million output tokens, flat across the whole context window. Cached input is $0.075 per million. Data Studios reported slightly lower prices of $1.413 and $4.24 in Alibaba's other regions, including US (Virginia), Frankfurt, Tokyo, Hong Kong and Beijing.
Qwen3.8-Max listed at $2.00 and $6.00 in the Singapore region, so the new flagship is 25% cheaper there, according to Data Studios. Its input price matches Google's Gemini 4 Hydrogen, Google's fast model, though Hydrogen charges $7.50 per million output tokens. For how these per-token prices work, see why AI prices are quoted per million tokens.
Alibaba also says the model does the same work with fewer tokens. On its Work360 test of professional tasks, Alibaba reported that the average cost per task fell from $4.43 with Qwen3.8-Max to $1.67, according to Digital Trends and DataCamp. BigGo Finance, working from yuan figures, put the drop at 58%.
How it scores
On Alibaba's chart, Qwen4-Max scores 48.5% on Terminal-Bench 4.0, a test of agents working in a computer terminal, up from 38.9% for Qwen3.8-Max. On the same chart, Anthropic's Claude Opus 5.5 scores 59.6%, OpenAI's GPT-6.1 Sol 56.1% and Moonshot's Kimi K3.1 67.2%, DataCamp reported. Alibaba's model leads that group on its own Agents' Last Exam result, at 46.7 against 43.4 for Opus 5.5.
Outside testers place it below the leaders. DataCamp reported that Artificial Analysis ranks Qwen4-Max ninth on its Intelligence Index with a score of 43, and Vals ranks it 11th of 38 models. Vals measured 48.48% on Terminal-Bench, matching Alibaba's number. Vendor charts pick the tests and settings, so the independent rankings are the better guide; our benchmark explainer explains why.
Why it matters
Qwen4-Max gives developers a large, agent-focused model from a Chinese lab at a lower price than its predecessor, but it is not the open model many Qwen users are used to. With independent rankings placing it in the top ten rather than at the top, its main pitch is cost per task, which adds to the price pressure on US labs after DeepSeek's latest API changes.
Sources
- Qwen4-Max: Features, Benchmarks, and Pricing, DataCamp, November 16, 2026
- Qwen4-Max is China's newest flagship AI model, and it makes GPT and Claude look expensive, Digital Trends, November 16, 2026
- Qwen4-Max API pricing: $1.50 input, $4.50 output per 1M tokens, 1M context, Data Studios, November 16, 2026
- Alibaba Releases Qwen4-Max: 2.4 Trillion Parameters, Per-Task Cost Slashed by 60%, BigGo Finance, November 16, 2026
Written by an AI correspondent under standards set by editor Jacob Knight. Spotted an error? Tell us and we will fix it.