Baidu Comate Leads AI Code Generation Market According to Industry Reports

On September 25th, the international authoritative evaluation agency Sullivan, in collaboration with TouBao Research Institute, released the "2024 AI Code Generation Market Observation Report - Milestone" annual report. The report comprehensively assessed the competitive performance of six Chinese AI code generation vendors including Baidu, Alibaba, Tencent, iFlytek, SenseTime, and ZhiP AI across ten dimensions. The results showed that Baidu's Wenxin KuaiMa (Baidu Comate) ranked first in eight important dimensions including technical foundation, application implementation, and product maturity, becoming the product with the most first-place rankings among all evaluated vendors and taking the top position. On the same day, domestic evaluation agency SuperCLUE also released the "Chinese Native Code Assistant Evaluation Benchmark Ranking," where Baidu's Wenxin KuaiMa achieved a total score of 87.55, ranking first among domestic AI code products.

The Sullivan report evaluated from four stages and ten major dimensions, including initial stage, exploration stage, development stage, and maturity stage, mainly examining the technical research and development capabilities, market performance, and future development potential of AI code generation products.

The report pointed out that as a leading enterprise in China's AI code generation, Baidu's Wenxin KuaiMa relies on the Wenxin large model and Baidu's programming big data to achieve intelligent assistance throughout the entire process of "thinking, writing, and modifying," significantly improving coding efficiency and releasing "ten times" software productivity. This allows developers to be freed from tedious code and focus on innovation, accelerating product iteration and upgrades. The report also highlighted three major highlights of Wenxin KuaiMa: over 20 years of technical accumulation, cost reduction and efficiency improvement, and simplicity and ease of use.

In the initial stage, the report examined from two dimensions: technical foundation and conceptual prototype. The amount of code data, deep learning, and natural language technology are the starting points for AI code generation competition. Wenxin KuaiMa, with the advantages of the Wenxin large model, combined with a large amount of code data and natural language dialogue data in the development environment, as well as years of software development experience accumulation, gives it a leading advantage.

In the exploration stage, the implementation of functions such as code completion, prediction, and refactoring, as well as scenario implementation, are key to this stage. The report believes that Wenxin KuaiMa, combined with Baidu's twenty years of programming on-site big data, continuously iterates in code generation, code explanation, generating comments, and code optimization, comprehensively covering the entire research and development process. Its mature product functions and application capabilities effectively meet market and enterprise needs.

In the development stage, the report mainly examined the product's technical competitiveness, marketization, and application implementation capabilities. Technically, Wenxin KuaiMa not only provides basic functions but also has highly scenario-based capabilities, which can seamlessly connect to enterprise private domain knowledge, including business interface documents, product requirement documents, test case documents, etc. In terms of implementation, Wenxin KuaiMa has rapidly penetrated into multiple industries such as automotive, finance, Internet, and mechanical manufacturing.

In the maturity stage, the report mainly examined whether the product has achieved near-complete automatic programming and whether it has become an industry benchmark. Sullivan believes that Wenxin KuaiMa can deeply understand the research and development process and provide strong support in various links from demand analysis, architecture analysis to testing and operation and maintenance. At the same time, Wenxin KuaiMa can automatically complete many repetitive and low-creative tasks, such as debugging, generating code comments, and writing repetitive code, freeing up developers' time so they can focus more on the implementation of core logic and innovative work.

Public information shows that Wenxin KuaiMa was officially launched externally in June last year and has now been updated to version 2.5, covering more than 200 programming languages including Java, Python, and C++, and supporting more than 10 mainstream IDEs such as JetBrains and VSCode.

Within Baidu, 85% of engineers have been deeply using Wenxin KuaiMa, with the proportion of newly generated code reaching 30% and code adoption rate exceeding 46%. In addition, Wenxin KuaiMa has also been applied to tens of thousands of enterprise customers such as Geely Automobile, SF Technology, Founder Securities, Hua Nong Insurance, Himalaya, Tongcheng Travel, MINISO, and Shanghai Mitsubishi Elevator, covering industry fields such as automotive, finance, logistics, Internet, mechanical manufacturing, software services, and integrated circuits.

In addition to the Sullivan report, Baidu's Wenxin KuaiMa also ranked first among domestic manufacturers in the SuperCLUE evaluation benchmark list. The evaluation agency SuperCLUE released the first "Chinese Native Code Assistant Evaluation Benchmark List" on September 25th, evaluating the capabilities of two overseas AI code products and five mainstream domestic AI code products from four first-level dimensions and eleven second-level dimensions. The results showed that Baidu's Wenxin KuaiMa achieved a total score of 87.55, ranking first among domestic AI code products, and scored higher than other domestic products in code generation and rewriting, code understanding and analysis, and code optimization and repair, being in a leading position.

At present, AI code generation has become one of the most widely applied scenarios for global AI generation landing. AI intelligent coding assistants can effectively help enterprises reduce development costs, shorten development time cycles, and improve project acceptance pass rates.

Sullivan predicts that by 2028, the scale of China's AI code generation market will grow to 33 billion yuan, and in the future, AI code generation will penetrate into more industries and fields.