Alibaba's flagship model, Qwen4-Max, has reduced the average cost per task from RMB 26.58 to RMB 11.16, a decrease of 58%.
Citi analyst Alicia Yap's team maintained a "Buy" rating on Alibaba (BABA) in a report on November 16, with a price target of $193, representing a 109.3% upside from the intraday price of $92.21 on November 16.
The row believes that Qwen4-Max signifies a shift in the large model competition from merely accumulating parameters to achieving enterprise-level economic efficiency and deploying 24/7 agent workflows.
The activation parameter is only 7.1 billion, and token consumption has been reduced by nearly half.
On November 16, Alibaba officially launched Qwen4-Max, its flagship mixture-of-experts (MoE) foundational model based on the new Qwen4 architecture. The model has a total of 2.4 trillion parameters, but activates only 71 billion parameters per generated token, while supporting a 1 million token context window and native visual processing.
Architectural changes directly translate into improved inference efficiency. According to the Citi report, Qwen4-Max introduces Sparse Attention (QSA), Gated Residual (GR), and 328 billion N-gram embeddings, significantly reducing computational overhead for long-task inference.
In practical tests, Qwen4-Max shows a significant reduction in token consumption compared to its predecessor, Qwen3.8-Max. Token consumption decreased by 24% to 35% under Work360’s high-inference load, and by 48% to 49% in the Terminal-Bench 4.0 terminal programming test.
The saved tokens directly translate into real savings for enterprise clients. According to Citi data, the average cost per task for Qwen4-Max has dropped to RMB 11.16, a reduction of approximately 58% compared to Qwen3.8-Max’s RMB 26.58.
The API pricing also maintains high cost-effectiveness. Qwen4-Max is now available for invocation on the Tongyi Qianwen platform, with domestic input and output prices at RMB 10 and RMB 30 per million tokens, respectively, and international prices at $1.5 and $4.5 per million tokens.
From writing code to financial due diligence, models are taking over complex, long-term tasks.
After reducing costs, Alibaba is deploying Qwen4-Max into enterprise production environments that require multi-step collaboration.
In the web development evaluation CodeArena, Qwen4-Max scored 1750 points, ranking sixth globally. In the long-term, high-value task test Agents' Last Exam v1.1, its score increased from 27.9 in the previous generation to 46.7.
Citibank outlined four practical scenarios demonstrated by Alibaba, all centered on using agents to replace manual execution of lengthy processes:
First, industrial and vertical industry coverage. Alibaba’s Work360 scenario framework covers 26 primary vertical industries, 182 secondary industries, and 721 tertiary sub-sectors. In the automotive manufacturing sector, this model has already begun supporting production scheduling, component design, and process validation.
Next is 3D content and game development. By integrating with Blender, Qwen4-Max supports architectural design, material assignment, and lighting rendering, and can create scenes and program gameplay logic in Unity, Unreal Engine, and Godot using natural language commands.
In financial and legal contexts, Qwen4 can access 90 MCP tools covering financial reports, interest rates, and derivatives data for A-shares, Hong Kong stocks, and U.S. stocks. The model can perform multi-factor backtesting and equity-bond comparisons, directly outputting reports in Excel or PPT format, and assist with legal due diligence for corporate IPOs and multi-source clinical decision-making.
To support extended unattended operation, Qwen4-Max introduces Live Delivery, enabling agents to dynamically adjust output based on user feedback during execution.
Cloud inference demand becomes the core growth driver, with segment valuation seen as high as $193.
Citibank believes that the accelerating adoption of enterprise agents and Model-as-a-Service (MaaS) will be the core driver of Alibaba Cloud's future revenue growth. As task execution costs decline by 58%, more enterprises are willing to integrate their software development, manufacturing scheduling, and financial research workflows with Alibaba Cloud's interfaces, driving sustained growth in demand for inference computing power.
In its valuation approach, Citibank used the sum-of-the-parts (SOTP) method to reprice Alibaba. The bank assigned a P/E ratio of 10x to Alibaba’s Taotian E-commerce Group’s net profit for fiscal year 2027, a P/S ratio of 7.0x to its AI cloud and computing services revenue for fiscal year 2027, a P/S ratio of 5.0x to its AI lab and application revenue, and a P/S ratio of 0.2x to its remaining businesses, while applying a 30% discount to its net cash position.
After summing up all business segments, Citigroup arrived at a fair value of $193 per share for Alibaba, with an expected dividend yield of 2.2%, projecting a total return of 111.5%.
The report also highlighted six key downside risks: underperformance in the execution of the new retail strategy, reduced profit margins due to increased strategic investments, slowing growth in user traffic and online gross merchandise volume (GMV), integration risks associated with newly acquired businesses, trade tensions between China and the U.S. and macroeconomic volatility, and regulatory risks stemming from merchant quality control issues.
Institutional ledger

This article is a compilation and interpretation by Chaoxiang Research of third-party brokerage research reports. The ratings, price targets, earnings forecasts, and related judgments cited herein are the views of the brokerage's analysts and represent the positions of their respective institutions; they do not reflect the views of Chaoxiang Research nor constitute any investment advice.
The market carries risks; make decisions independently. This article should not be used as a basis for buying or selling any securities.



