Author | Chen Junda, Editor | Mo Ying
The popularity of the AI coding track in China is evident to all. Over the past year, large model vendors have intensively launched AI coding products, and startups have flocked to this emerging market. Companies such as Alibaba, ByteDance, Tencent, Baidu, Zhipu AI, and Moonshot AI have successively increased their investments in the AI coding field, leading to the rapid emergence of a large number of products. Meanwhile, the dimensions of competition in AI coding are continuously enriching. Model capability is merely the foundation; development experience, scenario coverage, user scale, and enterprise deployment capabilities are becoming key metrics for evaluating product competitiveness. However, for a long time, the outside world has lacked authoritative data references regarding the actual market share and leading players in the domestic AI coding market.
Which products have truly gained recognition from developers and enterprises? And who has taken the lead in achieving scaled applications?
Recently, the latest "China AI Coding Market Share, 2025" report released by the internationally renowned market research firm IDC provided the first answer to this market. The report shows that Alibaba's Qoder, with a 47.6% market share, has secured the first place in revenue share, accounting for half of the domestic AI coding market and even exceeding the combined market share of the second to fifth places. Meanwhile, the total user base of Qoder's full product line has surpassed 5 million, making it the most popular AI coding tool in China.
In June this year, the market research firm Gartner released the "Magic Quadrant for Enterprise AI Code Agents 2026" report, with only 12 companies globally shortlisted. Relying on Qoder, Alibaba Cloud has entered the "Challengers" quadrant for three consecutive years and is also the only Chinese company to enter this quadrant.
Why Qoder? What industry changes are reflected behind these report cards?
01.Entering the Agentic Coding EraQoder Hits the Right Direction of Evolution
For many developers, Qoder taking the first place in the market is not an unexpected result. On social media platforms, the discussion heat about Qoder among developers is quite high. Some netizens shared their experience after migrating from OpenAI products to Qoder, stating that Qoder can already handle most daily development tasks, and the price of around 70 RMB per month is also very cost-effective.
Under a discussion about which AI coding product is the most user-friendly in China, many netizens spontaneously recommended Qoder. Reviews such as "user-friendly," "cost-effective," and "stable" have become repeatedly appearing keywords.
Qoder is also one of the fastest-iterating AI coding products in China. Since August last year, Qoder and its predecessor, Tongyi Lingma, have basically maintained a pace of almost weekly updates, continuously introducing the latest domestic flagship models while steadily optimizing the AI coding experience. In May this year, Qoder 1.0 was officially launched. One of the most popular new features in this generation of products is the Quest independent window. This window integrates capabilities such as task management, status tracking, artifact review, and knowledge invocation. Users only need to input requirements, and the entire process of execution and verification can be autonomously completed by AI within the workspace.
Qoder 1.0 also supports cross-project multi-task parallel processing. Users can grasp project progress within a single screen without switching windows, which improves the development efficiency of complex projects.
This working mode of Qoder aligns with the current Agentic coding trend. AI expert Andrej Karpathy once lamented that the efficiency bottleneck of today's AI coding tools is actually the throughput of "humans" themselves. Therefore, an excellent AI coding tool, in addition to producing high-quality code, also needs to help users quickly understand what the Agent is doing, control the task execution process, and verify and optimize the final results. Targeting the enterprise market, Qoder 1.0 has also launched many practical features. Regarding security, which is the most concerning issue for enterprises, Qoder has built a three-layer security architecture covering the pre-event enterprise policy layer, in-event security runtime, and post-event audit governance layer, controlling security through various means such as regular expression validation, semantic diff, cross-file analysis, and CI/CD interception. How to accumulate enterprise experience and knowledge is a mandatory question for AI coding products in enterprise scenarios.
Currently, Qoder has integrated originally scattered functions such as memory, Repo Wiki, and knowledge cards into a unified knowledge engine, enabling team-level knowledge sharing. When executing tasks, Qoder can independently invoke information such as team specifications, historical decisions, module relationships, coding standards, and tech stack knowledge. With the assistance of this information, Qoder can produce high-quality code more efficiently. Actual measurement data shows that after the launch of the knowledge engine feature, the user code retention rate has increased by 11%, input Token consumption has been significantly reduced by 40%, and the number of dialogue turns has decreased by 33%.
Through a dedicated plugin/skills marketplace, Qoder also enables enterprises to build their own exclusive skill libraries, allowing these workflows and toolchains to flow freely within the enterprise. Supporting the above product innovations is Alibaba's continuous iteration in underlying models and Harness technology. Currently, one of the main models behind Qoder is Alibaba's Qwen 3.7-Max. Upon its release, this model achieved multiple SOTAs in the coding field. In the SWE-Bench-Verified benchmark test, its score of 80.4 is already comparable to first-tier industry models such as Opus 4.6 Max and DeepSeek-V4.
Of course, Qoder has also adopted an open strategy, keeping the latest top domestic and international models such as GLM, DeepSeek, and Kimi online, and handing the choice over to users. Harness is the key to fully leveraging model capabilities. Today, Qoder has abandoned the traditional chat dialogue architecture and upgraded to a structured task runtime. From a specific design perspective, Qoder's upgraded runtime achieves Workspace binding, allowing tasks to be created from the source project and run in the bound execution environment. Artifacts, Reviews, and Commits can all be mapped to clear delivery targets. Its Artifact pipeline structures the Agent's execution process into a reviewable artifact chain, with clear ownership and status at every step. Once the task boundaries are stabilized, the completion rate of complex tasks improves by over 60%.
02.AI Coding Is Breaking Away from IDEsQoder's Forms Continue to Enrich
Reviewing Qoder's recent upgrades, it can be found that another main thread is the continuous enrichment of product forms. In the past, when people mentioned AI coding, they often thought of code completion in editors or the chat window on the left side of the IDE. Developers open the IDE, ask questions, and wait for AI to answer. However, as Agent capabilities continue to mature, this interaction method has begun to show its limitations. True software development has never been a process of sitting fixedly in an IDE.
Developers debug programs in the terminal, consult documents in the browser, handle urgent needs on their mobile phones, and run automated tasks in the cloud. If AI is always restricted to an editor window, the development links it can participate in remain limited. Currently, Qoder has formed a product system covering multiple forms such as Desktop, JetBrains plugins, CLI, Mobile, Cloud Agents, and QoderWake. For a large number of senior developers, the command line remains the most efficient working environment.
Compared to frequently switching windows, they are more accustomed to directly completing code management, deployment, and automated script execution through the Terminal. The CLI form of Qoder allows developers to directly invoke the Agent in the command line to complete tasks such as project setup, code modification, and task execution, integrating AI into the original development process rather than requiring developers to change their long-formed work habits.
The emergence of the mobile end also reflects Qoder's understanding of the changes in development scenarios. Traditional AI coding products are mostly limited to the PC end, and as Cloud Agents begin to have continuous running capabilities, a complete development task often takes tens of minutes or even hours to complete. After developers leave their computers, the task continues to advance in the background. The value of Qoder Mobile lies precisely here. Developers can check the Agent's execution status in real-time via their mobile phones, review generated code, append new instructions, and even remotely create Cloud Agent environments, achieving "humans leave the computer, Agents continue to work."
If CLI and Mobile are more about the expansion of interaction entry points, then Cloud Agents represent the extension of AI Coding capabilities from personal tools to infrastructure. In the past, AI coding mainly served individual developers, while Cloud Agents further open Agent capabilities to enterprise scenarios. Enterprises can invoke managed Agents through APIs without building complex running environments, integrating capabilities such as code generation, test verification, and automated development into the existing R&D system, further improving the automation level of the software development process.
QoderWake is a more exploratory form in the Qoder product matrix. It is designed as a virtual R&D partner with job responsibilities, long-term memory, and task boundaries. Developers can assign tasks, track progress, and continuously drive task iterations just like managing real team members.
From IDE plugins to the command line and mobile end, and then to cloud Agents and virtual R&D partners, Qoder is gradually covering different links in software development, enabling Agents to move from assisting developers in completing single-point tasks to further participating in the complete process of requirement understanding, code writing, test verification, and continuous delivery.
03.The Fierce Battle in AI Coding ContinuesHow Much Weight Does Qoder's "First Place" Carry?
The AI coding track where Qoder operates has become one of the most watched and fastest-growing tracks in the competition for large model applications, and it is also the main battlefield that no AI player can ignore. Unlike many AI application scenarios, coding tasks naturally possess high verifiability. Whether the code can run, whether tests pass, and whether development efficiency improves can all be measured through objective indicators. Every improvement in model capabilities can also be quickly reflected in development efficiency, code quality, and development experience.
Therefore, AI coding is not only an important scenario for the landing of large model capabilities but has also become a "touchstone" for testing model reasoning capabilities and Agent capabilities. More crucially, at the commercialization level, AI coding can be said to be one of the large model applications with the clearest commercialization path currently. Developers themselves are a group highly sensitive to efficiency and willing to pay for high-quality tools; for enterprises, the R&D team is often the department that invests in AI applications the earliest and gets the most direct returns. The efficiency improvements brought by AI coding, such as code generation, automated debugging, and R&D process optimization, are easier to be quantitatively evaluated, and the return on investment is also easier to calculate. Therefore, whether individual developers or enterprise R&D teams have shown a strong willingness to pay for AI coding tools that can truly enhance productivity.
Tools with technological advantages and better product experiences have even become "scarce resources" that developers scramble to snatch up. In such a high-speed growing and fiercely competitive emerging market, Qoder taking the first place in domestic market share speaks for itself. Currently, Qoder has served over 5 million global users and hundreds of thousands of enterprises, and has entered the scenarios of leading enterprises such as FAW Group, CITIC Securities, and AsiaInfo Technologies, indicating that Alibaba's AI capabilities are further penetrating into the core production links of enterprises from the underlying models and infrastructure. At the same time, it can also be seen that AI coding tools possess strong user stickiness.
Development tools themselves have high migration costs. After long-term use, developers will accumulate a large amount of personal habits, project context, codebase information, and workflow configurations, and the tool's understanding of the user's development methods will continue to deepen, forming a positive feedback loop of "the more you use it, the better it understands you." Qoder's nearly 50% market share represents that it has already possessed a certain scale of developer ecosystem and user mindshare advantage. As users' usage habits continue to deepen, and development process data and scenario experience continue to accumulate, Qoder's leading advantage may further expand.
04.Conclusion: AI CodingHas Become a Battle for the Entry Point
Reviewing the technological development of the past few decades, Windows for PCs, iOS and Android for the mobile internet, and AWS for cloud computing have all changed the industrial landscape by mastering the developer entry point. Today, AI coding tools are becoming new entry-level products in the AI era. Because almost all developers are accelerating their product iterations through AI coding tools, mastering the developer entry point means connecting to the source of the future AI application ecosystem. For Qoder, being first in the market is not just about leading in share; it more importantly means that on the eve of the explosion of AI applications, it is seizing a strategic position similar to an "operating system level."