Qwen4-Max Launches with 25% Fewer Active Parameters and 62% Lower Task Costs

iconKuCoinFlash
Share
AI summary iconSummary
Alibaba has launched Qwen4-Max, a new AI model featuring 25% fewer active parameters and 62% lower task costs. The model has a 2.4 trillion parameter backbone, 710 billion active parameters, and 328 billion N-gram embeddings. It achieved a score of 52.69 on the Work360 benchmark and now costs $1.67 per task. Traders monitoring altcoins to watch may note the model’s enhanced efficiency. The Fear and Greed Index remains a key metric for gauging market sentiment amid these AI advancements. Qwen4-Max is available via API at reduced pricing.
ME AI News, Dongcha Beating AI Brief: Following last week’s preview release, Alibaba has officially launched its next-generation flagship model, Qwen4-Max, disclosing its full technical architecture, evaluation results, and API pricing. Qwen4-Max retains the 2.4-trillion-parameter backbone of its predecessor and introduces an additional 328-billion N-gram embedding parameter. By redesigning the attention mechanism, residual connections, and phrase embedding layers, the number of backbone parameters activated per generated token has decreased from approximately 95 billion to 71 billion—a reduction of about 25%. The model natively supports visual understanding and a 1-million-token context window. Official evaluations show Qwen4-Max scored 52.69 on the Work360 professional benchmark, up from 47.74 for the previous version. This proprietary Alibaba benchmark covers professional tasks in finance, law, manufacturing, and more. On the Terminal-Bench 4.0 programming test, the new model reduced token consumption by approximately 48% to 49% compared to its predecessor. The average cost per task on Work360 dropped from $4.43 to $1.67—a 62% reduction. Alibaba also demonstrated a research case: Qwen4, integrated with an automated experimentation platform, autonomously ran for 48 hours to screen novel coolant materials for data centers from 2.7 million candidate molecules. After four rounds of molecular design, synthesis, and testing, it successfully produced physical samples of validated candidate molecules. However, official data proving its actual cooling performance surpasses existing materials has not yet been disclosed. In an automotive manufacturing case, Qwen4-Max assisted in part modeling, machining path planning, and simulation validation. The manual time required for a machining programming task was reduced from approximately 475 minutes to just 24 minutes—a 95% decrease. The model can also use Blender, Unreal Engine, and Unity to create 3D scenes, playable games, and iteratively refine them based on runtime feedback. Qwen4-Max is now available via official API with OpenAI and Anthropic-compatible interfaces. Pricing is set at ¥10 per million input tokens and ¥30 per million output tokens—about 17% cheaper than the previous version; cached input pricing has dropped from ¥1.5 to ¥0.5. Alibaba confirmed that smaller models in the Qwen4 series will be open-sourced in the future. (Source: MLion)
Disclaimer: The information on this page may have been obtained from third parties and does not necessarily reflect the views or opinions of KuCoin. This content is provided for general informational purposes only, without any representation or warranty of any kind, nor shall it be construed as financial or investment advice. KuCoin shall not be liable for any errors or omissions, or for any outcomes resulting from the use of this information. Investments in digital assets can be risky. Please carefully evaluate the risks of a product and your risk tolerance based on your own financial circumstances. For more information, please refer to our Terms of Use and Risk Disclosure.