Recently, ByteDance's AI-native programming tool Trae released its "2025 Product Report," providing a comprehensive overview of the growth and user insights from the past year. As a leading AI Coding platform in China, Trae has been iterating at an astonishing speed and has become an indispensable "intelligent partner" for developers worldwide. According to the report data, this tool is profoundly reshaping the programming paradigm and driving AI from auxiliary completion to full-cycle autonomous development.

User Base Explosion: 1.6 million monthly active users, over 6 million total registrations

As of the report release, Trae's total number of registered users has exceeded 6 million, covering nearly 200 countries and regions, with monthly active users surpassing 1.6 million. This figure stands out among professional programming tools, particularly ranking at the forefront in the domestic market. User activity is extremely high, with an average of nearly 5 active days per week, equivalent to "full attendance on workdays," while international paid users even approach "full attendance throughout the natural week." Throughout the year, users initiated nearly 60 million sessions and 500 million queries, generating nearly 100 billion lines of code for global developers. The daily token consumption has increased by nearly 700% in the past six months.

TRAE 月活破160万！借着数据聊 AI 编程生态Trae 发布了他们的年度开发者报告。也可以看作 (1).jpg

Performance Optimization Significant: Latency and Memory Issues Substantially Alleviated

In 2025, Trae underwent rapid iteration from Agent 1.0 to the official SOLO version, significantly improving the product experience. The report indicates that code completion latency was reduced by 60%, and memory usage dropped by 43%, effectively addressing early user pain points. The penetration rate of the SOLO mode reached 44% (international version), and after the Chinese version was launched, three out of every ten developers used it, marking the evolution of AI from single-point assistance to complex task handling.

TRAE 月活破160万！借着数据聊 AI 编程生态Trae 发布了他们的年度开发者报告。也可以看作 (2).jpg

Usage Scenario Insights: Bug Fixing and Code Generation Dominate

In IDE mode, the most common tasks for users are bug fixing (accounting for about 40%) and code generation (about 30%), which are the areas where AI models perform best and developers have the highest trust. In contrast, features such as repository management, environment management, and code optimization are used less frequently, reflecting users' cautious attitude towards AI in complex engineering tasks.

The distribution of programming languages shows that Vue is first, followed by Python, JavaScript, HTML, Java, TypeScript, etc., indicating that front-end web development remains the main battlefield for AI coding.

TRAE 月活破160万！借着数据聊 AI 编程生态Trae 发布了他们的年度开发者报告。也可以看作 (3).jpg

Clear Mode Preferences: Builder Dominates, Hybrid Usage Becomes Common

Trae supports a wide range of AI collaboration modes, with the Builder mode being the most favored by users (accounting for 80%), followed by SOLO Coder (40%). More than 80% of users adopt multiple intelligent agents, indicating that developers have mastered the advantages and limitations of different modes. At the same time, the platform supports over 11,000 MCP (Model Connection Protocols), enabling deep integration with external tools, further expanding the application of AI in automated scenarios.

TRAE 月活破160万！借着数据聊 AI 编程生态Trae 发布了他们的年度开发者报告。也可以看作.jpg