Tencent discloses R&D progress for the first time: AI generates 50% of new code, and the level of R&D automation has increased by 67% year-on-year

Artificial intelligence is evolving from a cutting-edge concept into a core driving force for internal innovation among China's tech giants. Tencent's latest "2025 Tencent R&D Big Data Report" shows that AI has been deeply embedded in its vast R&D system, not only accelerating the software development process but also significantly enhancing overall efficiency and product delivery speed, becoming a key engine for this tech giant to maintain long-term competitiveness.

The most striking data in the report is that AI has become a daily work partner for Tencent engineers. Over 90% of engineers are using AI programming assistants, and 50% of the new code is generated with AI assistance. This transformation has directly driven an overall improvement in R&D efficiency of over 20%, with Tencent's level of R&D automation increasing by 67% year-on-year, saving 5.3 million manual operations each month. This demonstrates that Tencent's long-term investment in AI is accelerating its conversion into business value.

Behind these advancements is Tencent's massive investment in R&D and organizational scale. The report shows that R&D personnel account for 76% of Tencent's total workforce, meaning that three out of every four employees are engaged in R&D work. For a company that adds over 325 million lines of code each month, the efficiency gains brought by AI mean a direct enhancement in product iteration speed and market responsiveness, which is crucial for investors assessing its future growth potential.

AI Fully Integrated into R&D Processes, Dual Efficiency in Coding and Review

The report elaborates on how AI has permeated various stages of the software development lifecycle. Supported by its self-developed mixed Yuan model, AI is no longer just an auxiliary tool but is deeply involved in core tasks such as coding, code review, and testing.

Data shows that with 50% of the new code completed with AI assistance, the average coding time for engineers has been reduced by 40%. This means developers can devote more energy to more creative and complex tasks. In terms of code quality control, AI's participation rate is as high as 94%, acting as an "AI quality inspector." It conducts pre-reviews before human engineers intervene, directly identifying and adopting fixes for 28% of code defects, driving a 44% increase in the effective problem detection rate during code reviews, thus building the first line of defense for software quality.

R&D Efficiency Platform Supports Automation Leap, Delivery Speed Significantly Accelerated

The large-scale implementation of AI relies on the support of underlying R&D platforms. The report points out that as the WeDev R&D efficiency platform deepens its integration into R&D practices, Tencent's level of R&D automation has increased by 67% year-on-year, saving 5.3 million manual operations each month.

The efficient platform support has led to a significant increase in delivery speed. By 2025, Tencent will complete an average of 16,000 demands per day, a year-on-year increase of 25%, with the average completion time shortened by 12 hours. Among them, the AnyDev cloud R&D platform has drastically reduced the environment preparation time from one day to just one minute. In terms of code quality, the combination of automated tools and AI technology has repaired over 5.4 million code defects and security vulnerabilities throughout the year, with the average bug resolution time shortened by 8 hours, achieving "early detection, early repair."

From WeChat to Games, AI Enhances Efficiency Across All Business Lines

The efficiency improvements brought by AI and platformization have been validated across Tencent's major business lines, translating into tangible business results. The report states that 81% of the R&D teams have achieved full-process efficiency improvements relying on the WeDev platform.

Specifically, the WeChat backend team has reduced compilation time by 50% through a distributed compilation toolchain; the demand delivery cycle for WeChat Pay has been shortened by 31%, and the quality of releases has improved by 14%. In Tencent's other major pillar business, the gaming sector, the automation rate of art production has reached 95%.

At the same time, 65% of the new code in Tencent Cloud comes from the AI code assistant Codebuddy, with the bug rate per thousand lines of code reduced by 31.5%; the iteration efficiency of Tencent Advertising has doubled, with 90% of version releases achieving full-process automation.

Continuous High Investment and Open Source Strategy Highlight Technological Layout

Behind these R&D achievements is Tencent's continuous high investment in R&D and a clear technological strategy. According to its second-quarter financial report, R&D expenditure for the quarter reached 20.25 billion yuan, and cumulative R&D investment has exceeded 379.5 billion yuan since 2018.

Supporting Tencent's AI system is its long-term technological investment and open strategy. Tencent's Hunyuan large model continues to iterate and fully embraces open source, with its image model "Hunyuan Image 3.0" ranking first in global user blind testing on the international large model evaluation platform LMArena.

In the external open-source ecosystem, Tencent's open-source projects on GitHub have accumulated over 520,000 stars, ranking among the top ten globally. The company has also contributed a series of widely used open-source tools in the industry, including the Tinker hotfix framework, WeTest automation testing platform, RapidJson parsing library, and Kona JDK.

In terms of programming language preferences, C++, Go, Python, and Java remain the mainstays of Tencent's R&D system. Go language is widely used in backend services due to its high performance and simple design, while Python has become the preferred language in AI and large model projects, reflecting the continuous evolution of Tencent's tech stack.