Alibaba's Qwen App released a major update on August 7th, integrating an office assistant for consumer AI, scheduled tasks, and access to the flagship Qwen3.8-MAX model into one version. For the first time in such a release, there is a clear pricing signal: core functions remain free, but the new office assistant will operate on a quota that requires payment upon exhaustion.
This approach mirrors the model established by ByteDance with its Doubao in June, which introduced a three-tiered subscription ranging from 68 to 500 yuan per month. Tencent followed suit, offering its personalized plans in the range of 99 to 999 yuan. Alibaba itself had previously, on August 3rd, quietly launched the standard corporate subscription for Qwen Office at a price of 98 yuan per month.
Office workflows have become a segment where Chinese AI developers believe users are willing to pay. However, Qwen App is in a more complex position compared to its competitors. It was created as Alibaba's flagship consumer product, integrated with Taobao, Alipay, Fliggy, and Amap, and accompanied by a New Year marketing campaign worth 30 billion yuan, which boosted daily active users from 7 million to a peak of 73.5 million.
After this campaign concluded, the number of daily active users dropped by almost half. Session time per user also fell: from 6.3 minutes to about 3 minutes post-campaign, representing a 51% decrease. The cost to acquire one user was approximately 144 yuan per DAU, significantly exceeding the metrics of ByteDance's Doubao and Tencent's Yuanbao.
Internally, Qwen App also faces the Qwen Office product, which Alibaba publicly unveiled on August 3rd under the leadership of new DingTalk CEO Chen Yusen. This office product consolidated three previously competing agent products—QoderWork, Wukong, and MuleRun—into a single entry point. Since DingTalk focuses on enterprise sales and Qwen Office handles the performance roadmap, Qwen App must defend its unique consumer identity, which it has yet to establish.
Thus, the focus on paid office functionality is less about monetization and more about narrative building. Whether Qwen App can ensure sufficient quota usage to be considered a full-fledged office product will determine if it becomes the consumer interface for Alibaba's AI stack or just another chatbot in a saturated market.
The Alibaba Cloud Qoder coding platform has released Better Harness on GitHub—a three-layer open-source framework. This tool is designed to analyze the workflows of coding agents and ensure their continuous improvement.
Better Harness functions as a bridge between Harness Engineering practices and Loop Engineering, connecting development methodology with evaluation models and ready-to-run implementations. The tool can integrate with various coding agents, such as Claude Code, Codex, Qoder, and OpenClaw, creating a universal analysis layer applicable across different agent platforms.
The framework is built on a three-layer open architecture. The first layer covers Harness Engineering practices, including session management, Command Line Interface (CLI) integration, observability, rules, skills, MCP, memory, hooks, and automation templates. The second layer provides agent lifecycle evaluation models that transform engineering practices into verifiable questions, thereby limiting the relationship between evidence, evaluation, and conclusions. The third layer provides executable engineering implementations, allowing the first two layers to operate on real projects rather than remaining mere documentation.
Better Harness distinguishes between the presence of a capability and the effectiveness of that capability. The tool checks not just whether an agent has access to configurations, but whether it utilizes those capabilities and whether they support task completion. Every identified issue must be accompanied by traceable evidence, an assessment of user impact, a workload estimate, and verification methods. The tool automatically collects three types of evidence: Session Evidence obtained from observing direct interaction; Project Harness for repository configuration analysis; and Agent Customization for verifying agent-specific capabilities. An initial internal evaluation was conducted based on 30 real GitHub projects.
The release of this open-source tool reflects Qoder's strategy to establish Harness Engineering as a recognized discipline in AI-powered software development. By providing a standardized evaluation framework, Qoder aims to create an ecosystem where the workflows of coding agents can be objectively measured, compared, and improved across various agent platforms. This positions Better Harness as infrastructure for the emerging discipline of measuring and improving AI-driven development processes, similar to how testing frameworks created quality infrastructure for traditional software development.
Three major Chinese tech giants—Alibaba, Tencent, and ByteDance—have simultaneously focused their efforts on one strategic goal: the desktop office AI agent. According to a report by the analytics firm Analysys for the second quarter of 2026, covering 17 main desktop AI agents, the total number of visits in June 2026 reached 60.62 million. Of this, the three leaders—Alibaba, Tencent, and ByteDance—collectively accounted for 56.22 million of these visits, leaving only 4.4 million for all other competitors.
This data confirms that the platform economy applies to office AI agents, just as it did with browsers and search engines: users tend to install one primary desktop agent, which forms a natural oligopolistic dynamic.
In July 2026, all three companies restructured their AI agent portfolios. Alibaba merged three existing products—QoderWork (for local desktop operations), Wukong (for corporate security and permissions), and MuleRun (for complex workflow templates)—into a single product tentatively named Qianwen Office. This project is led by Chen Yusen, the new CEO of DingTalk, a technical founder born in 1992 whose startup Chaitin Technology was acquired by Alibaba Cloud.
Tencent reportedly moved its QClaw product center team to the Cloud Products Department 6 and integrated it with WorkBuddy under unified management. ByteDance, meanwhile, transformed Doubao into its flagship office agent brand.
The Chinese AI agent market is projected to grow from 80.4 billion yuan in 2025 to 696.8 billion by 2030. The most convertible segment consists of single-employee companies requiring AI assistance as an assistant, accountant, or designer. Nearly 60% of users are willing to pay because increased office productivity provides a direct return on investment: saved time translates into earned money.
Alibaba's strategy under Chen Yusen involves integrating three layers: QoderWork acts as the interaction layer for natural language desktop control, Wukong serves as the permission layer operating within corporate security boundaries, and MuleRun functions as the execution layer for reusable workflow templates. This approach is similar to Anthropic's methodology, where Claude Chat, Claude Code, and Claude Cowork utilize one base model under a unified interface. The goal of the unified product is to eliminate user confusion, which previously forced customers to choose between three incompatible Alibaba products.
Tencent's advantage lies in its deep integration with the WeChat ecosystem and cloud services. WorkBuddy already maintains a close connection with Tencent Docs, WeCom, and Tencent Meeting, offering a corporate workflow deeply embedded in China's dominant business communication ecosystem, WeChat. ByteDance Doubao leverages the company's experience in recommendation algorithms and content generation. This three-way competition boils down to which company can offer the most comprehensive office experience, from the start of user intent to task completion. The outcome of this battle will likely determine the direction of enterprise software in China for the next decade, as the desktop agent becomes the primary interface for knowledge workers interacting with digital tools, replacing the traditional office suite model with an AI-centric workflow paradigm.
