Author | Cheng Yanjing Editor | Hu Min
While the fervor surrounding Codex has not yet subsided, manual coding is being transformed by AI Coding.
In the early years, Microsoft introduced GitHub Copilot into daily development workflows, bringing AI into programmers' code editors for the first time; subsequently, companies like OpenAI further explored coding agents, enabling models to acquire software engineering capabilities such as understanding requirements and executing tasks.
Competition in the domestic market is also accelerating, with products such as Alibaba's Qoder, Zhipu AI's CodeGeeX, SenseTime's Raccoon, Tencent's CodeBuddy, and Baidu's Comate entering the arena one after another.
AI Coding is becoming the frontline for domestic large model companies competing for the developer ecosystem.
Recently, IDC released the "2025 China AI Programming Market Share" report, showing that Alibaba's agent programming platform Qoder has captured a market share of 47.6%, ranking first and surpassing the combined market share of Zhipu AI, SenseTime, Tencent, and Baidu.
In this most hardcore and challenging domain, how did Alibaba secure nearly half of the market?
01From "Code Assistant" to "Agent", AI Coding Enters a New Phase
To understand Alibaba's leadership, it is essential to first grasp the shifts taking place in the AI Coding landscape.
In the past, AI-assisted programming was largely confined to code completion and simple Q&A. Developers provided requirements, and AI generated code, but subsequent modifications, debugging, and engineering adaptation still required manual effort.
In the Agent era, the standards for AI Coding are being redefined. AI not only needs to generate code but also understand the tech stack, break down complex tasks, and execute, test, and fix within the software engineering workflow.
Given the Agent's capabilities in task planning, tool invocation, and continuous execution, AI is evolving from a "code assistant" to a "development participant."
From an enterprise perspective, AI Coding is also one of the directions with a relatively clear commercialization closed loop for large models at present.
Developers are a natural paying demographic, and enterprise R&D efficiency directly impacts product iteration speed and software costs. Compared to many AI applications still in the exploration stage, once AI Coding is integrated into the enterprise R&D process, it affects not only the individual efficiency of programmers but also has the potential to transform the entire software production paradigm.
Industry practices show that AI Coding tools can deliver a 30%-50% efficiency boost in certain R&D scenarios, helping development teams reduce repetitive work and devote more energy to architecture design and complex problem-solving.
This shift is also reflected in the talent competition. In the 2026 spring campus recruitment, Tencent expanded its technology positions by 36%, and ByteDance's "Jindouyun Talent Program" listed AI Coding as one of its eight key focus areas. Alibaba places even greater emphasis on it: AI-related positions account for over 60% of its 2026 campus recruitment, with AI positions at Alibaba Cloud and DingTalk reaching as high as 80%, and roles such as Qoder-AI Agent R&D Expert are directly open for recruitment.
Therefore, the competition in AI Coding is essentially no longer just a competition among programming tools.
It serves as both a touchstone for large model capabilities and a battle for the developer entry point in the Agent era.
Alibaba's Qoder securing the first place with a 47.6% share means it has already claimed a critical position in the battle for the developer entry point.
02How Does Alibaba Reconstruct AI Programming?
Reaping the 30%-50% efficiency dividend brought by AI Coding is far from as simple as stuffing a chat box into an IDE.
In a real software engineering environment, developers face not just writing code, but understanding complex legacy code, adhering to team standards, locating bugs, and ensuring the safe deployment of code.
According to market research from IDC and others, the reason Alibaba's Qoder can rapidly capture market share lies in driving the transformation of AI Coding from an "auxiliary tool" to "production infrastructure."
First, it is the shift from a "code assistant" to an "autonomous agent."
Early AI programming tools mostly adopted the Copilot model, where developers input code and AI handles completion. However, this model has obvious limitations when facing complex engineering: AI can generate code, but it cannot understand the complete project context, nor can it independently complete development tasks.
Take Microsoft's GitHub Copilot, for example. Initially taking on the role of code completion and assisted programming, Microsoft later began to drive Copilot's evolution into a more comprehensive software development assistant.
Looking at the approach of Qoder 1.0, it is about transforming AI from "completing a line of code" to "finishing a task." Through the Quest workspace, developers only need to propose target requirements, and the Agent can autonomously break down tasks, invoke tools, modify code, and complete testing and verification.
This will change AI programming interactions: in the past, it was "humans driving AI to write code," and in the future, it may become "humans defining goals, and AI executing engineering."
Supporting this change is the upgrade of Harness Engineering.
Traditional AI tools rely on chat context, whereas Qoder attempts to structure the Agent's execution process into a task runtime, ensuring that every action has a clear state and auditable results.
Second, it is the transition from a personal productivity tool to an enterprise knowledge engine.
Software development is essentially team collaboration, and large enterprises often accumulate massive code assets, technical specifications, and historical experience internally. Past AI programming tools served individual developers more, making it difficult to understand an enterprise's own technical system.
Qoder attempts to integrate enterprise knowledge scattered across code repositories and documents through capabilities such as Memory, Repo Wiki, and knowledge cards, enabling the Agent to understand the team's development standards, module relationships, and historical decisions.
According to official data, after the launch of the knowledge engine, the user code retention rate increased by 11%, input Token consumption decreased by 40%, and conversation turns were reduced by 33%, lowering the cost of repeated communication between developers and AI.
Finally, and also the critical threshold for enterprises to truly adopt AI Coding—security.
For large clients in finance, government, and enterprises, code is a core asset. If AI cannot guarantee permission isolation and data security, even the strongest model capabilities will struggle to enter the production environment.
In this regard, Alibaba's years of experience serving enterprise clients have formed an advantage. Qoder has established a security system covering pre-event, in-event, and post-event stages, featuring defense in depth from L1 to L4 through code detection, semantic analysis, and CI/CD interception. Mechanisms such as local data storage and high-risk operation interception guard the enterprise's security red lines.
Meanwhile, the enterprise edition, through the Plugin/Skill marketplace, knowledge management, and model permission configuration, enables different roles and code repositories to have matching AI capabilities. For large R&D teams, AI Coding is no longer just a personal productivity tool but has become part of the organizational R&D system.
In terms of product form, Qoder is also continuously expanding. From desktop IDEs, JetBrains plugins, and CLI, to mobile and Cloud Agents, and further to the QoderWake digital employee launched this year, Qoder is covering multiple scenarios from individual development to enterprise collaboration.
Behind this is Alibaba Group's long-term investment in AI infrastructure.
Eddie Wu previously announced an investment of 380 billion RMB over the next three years to build cloud and AI infrastructure; Alibaba Cloud halted its spin-off, continued to increase investment, and promoted the improvement of AI application levels across its business system.
Therefore, Qoder's market leadership is a concentrated manifestation of Alibaba's years of technological accumulation at the AI application layer.
03Moving Towards the "Development Entry Point" for the Next Decade
Of course, all engineering and productization innovations are ultimately built upon underlying model capabilities.
AI Coding, which tolerates no "hallucinations" whatsoever, extremely tests the underlying models. Qoder's ability to support complex enterprise-level Agent operations relies on the foundation of the Qwen3.7-Max model: while achieving SOTA in core evaluations such as SWE-Pro, it also delivered results on par with or even surpassing top overseas models in tests like Terminal Bench 2.0.
But the model is just an entry ticket. What truly creates the gap is the understanding of the next phase's form of AI Coding.
If you observe Qoder's recent moves, you will find it is attempting to redefine the relationship between humans and AI.
In May this year, Qoder launched Cloud Agents, a fully managed AI Agent runtime platform. By providing the Agent foundation, model services, and runtime environment in the cloud, enterprises can delegate tasks such as development, testing, and code review to cloud-based Agents, no longer confined to local development environments.
Previously, Alibaba launched the digital employee product QoderWake, which further expanded the role boundaries of AI.
Unlike traditional AI assistants, digital employees are endowed with clear job responsibilities, long-term memory, and work boundaries. Currently launched roles such as digital programmers and data analysts are transforming AI from a "passively responding tool" into a "virtual colleague capable of proactively taking on work."
This heralds that a brand-new organizational form of productivity is taking shape.
Looking at a deeper level, the leverage effect triggered by AI Coding far exceeds the programmer community itself.
The efficiency improvements brought by AI programming are changing the cost structure of the entire software industry and even the digitalization of various sectors. It transforms "AI for efficiency" from a slogan into measurable industrial value.
While overseas tech giants use Codex and Claude Code to compete for the developer entry point, Alibaba has proven another path with Qoder: by synergizing complete product forms, enterprise-grade security architecture, and cloud infrastructure, it upgrades AI Coding from a personal productivity tool to organizational productivity infrastructure.
The 47.6% share secured by Alibaba is more like conveying an industry signal: beyond competing in model intelligence, it is about capturing the developer's desktop.