Huawei Cloud has officially released the public beta version of its CodeArts code intelligence agent, providing developers and enterprises with an intelligent coding solution featuring engineering capabilities. The public beta is now available for free to developers, with quick start guides and comprehensive tutorials provided on its official website. Future iterations of Huawei Cloud CodeArts will focus on enhanced AI capabilities, more secure development environments, and more efficient engineering experiences, aiming to empower millions of developers and businesses with a new "AI Practitioner" development paradigm.

This version integrates a code large language model, an IDE, and an autonomous development mode, covering various AI coding technologies including code generation, development knowledge Q&A, unit test case generation, expert skills, codebase indexing, and specification-driven development. It incorporates open-source models like GLM-5.0 and DeepSeek-V3.2, alongside Huawei's self-developed models, and offers a dedicated model for HarmonyOS to improve developer efficiency and provide an intelligent coding experience.

Leveraging over two decades of Huawei's R&D experience and hundreds of billions of lines of code, CodeArts includes built-in Skills for high-frequency scenarios such as requirements management, system design, software development, compilation and building, testing and verification, release and deployment, and open-source and vulnerability management. It also supports rapid addition of custom skills for more convenient development.

For capability expansion, CodeArts integrates industry-leading models GLM-5.0 and DeepSeek-V3.2 with continuous training support, provides dedicated models for HarmonyOS and Ascend, and allows custom integration of third-party models. Regarding agent expansion, it offers a four-layer extension mechanism enabling developers to customize agents and sub-agents, expanding capability boundaries through MCP and Skills to meet diverse development scenarios and personalized needs.

Simultaneously, Alibaba is extending its AI initiatives into software development, with its cloud computing division launching low-cost programming tool subscription packages. On February 25, Alibaba Cloud's Bailian platform introduced a Coding Plan containing API services for four open-source models: Qwen3.5, GLM-5, MiniMax M2.5, and Kimi K2.5. Pricing starts at 7.9 yuan for the first month of the lightweight version (40 yuan thereafter) and 39.9 yuan for the professional version's first month (200 yuan thereafter). Alibaba Cloud stated this package primarily targets developer communities, allowing users to switch between mainstream AI programming tools like Qwen Code, Claude Code, and Cline. This move reflects the accelerated implementation of Alibaba's broader AI strategy.

Analysts suggest the open-source approach enables low-cost model access for users, while including multiple startup models in a single subscription further strengthens Alibaba Cloud's position as China's AI infrastructure platform. Cinda Securities notes that AI programming is reshaping core productivity methods, with large model technologies empowering programming tools. AI-driven automated programming and code generation enhance software development efficiency and automation levels. Key values of AI programming include: 1) improving software development efficiency and quality; 2) lowering technical barriers; and 3) accelerating project iteration cycles.

Major advancements in large models' programming capabilities are driving AI coding tools from Copilot assistance toward Agent models. Minmetals Securities indicates AI programming has become one of AI's earliest core applications, with multiple global tech giants launching related products. AI may first disrupt its own creators—AI programmers. Continuous progress in domestic open-source models is expected to form a crucial foundation for accelerated development of domestic IDEs.

Related concept stocks: Zhipu AI (02513): The recently launched GLM-5 large model is gaining widespread global attention. Technical reports detail its core architecture breakthroughs, demonstrating unprecedented capabilities in real-world programming tasks and comprehensively surpassing all previous open-source baseline models in handling end-to-end software engineering challenges. MINIMAX-WP (00100): On February 12, MiniMax launched its flagship programming model MiniMax M2.5. As the world's first production-grade model natively designed for Agent scenarios, its programming and agent performance rivals top international models like Claude Opus 4.6, supporting full-stack development across PCs, apps, and cross-end applications, particularly excelling in core productivity scenarios like advanced Excel processing, deep research, and PPT creation. Alibaba (09988): On February 25, Alibaba Cloud Bailian launched a Coding Plan featuring API services for four open-source models: Qwen3.5, GLM-5, MiniMax M2.5, and Kimi K2.5, with tiered pricing as described above.