SHENZHEN, China, March 10, 2025 (GLOBE NEWSWIRE) -- As 2025 kicked off, the release of the DeepSeek-R1 model by a Chinese AI startup sent shockwaves around the globe, delivering "OpenAI-level inference performance at just one-tenth of the training cost". The education industry quickly became the "eye of the storm" in this wave of AI democratization. Leading companies in various education sectors, including Xueersi, Youdao (by NetEase), Offcn Education, and Seewo, scrambled to integrate the DeepSeek-R1 large model, leveraging its capabilities to drive intelligent upgrades and innovations across both software and hardware. This surge is ushering the industry into a new era of AI-native education ecosystems.

I. Education Enterprises Rapidly Embrace DeepSeek, Unleashing a New Wave of Transformation with Its Chain-of-Thought and High Cost-performance Advantages
Since the official release of DeepSeek-R1 on January 20, 2025, it has seen rapid and sustained growth in popularity. Alongside a surge in consumer UV, leading education enterprises such as Youdao, Xueersi, Seewo, and Offcn Education have announced their adoption of DeepSeek. They are either integrating their existing AI-powered educational products with DeepSeek for capability and feature upgrades or using DeepSeek as a foundational model to launch new AI-driven educational applications and intelligent learning hardware. This strategic adoption positions them to seize the innovation opportunities brought by DeepSeek model in a short term.

Overview of Representative Educational Companies' Deployment on DeepSeek

Information Source: Compilation of public data

Compared to other general-purpose and domain-specific large models, DeepSeek has gained widespread favor and driven aggressive layout among leading education companies, thanks to several key characteristics:

• Deep Thinking Mode: DeepSeek's deep thinking mode outputs reasoning processes in natural language, making the thought chain behind learning and problem-solving visible. This is particularly beneficial in educational settings, as it helps illustrate problem-solving methods and processes, enables knowledge tracing, and guides students toward personalized thinking. This approach assists teachers and parents in education while mitigating the potential negative impact of AI+education models that previously focused solely on providing direct answers.
• Logical Reasoning Ability: DeepSeek has achieved comprehensive performance improvements across various types of reasoning tasks, making it especially effective in supporting the study of science subjects like mathematics and physics.
• Multimodal Capabilities: Following its multimodal training, DeepSeek large model has demonstrated enhanced performance in scientific tasks, complex reasoning, and mathematical coding. In educational contexts, its multimodal capabilities improve the understanding of text-image combined questions and enable the generation of diverse multimodal content, enriching AI-driven education applications.
• Open-Source and Cost Efficiency: As an open-source large model, DeepSeek provides access to its model architecture, parameters, and technical reports, allowing companies to build on it directly for secondary development. Furthermore, the cost of calling DeepSeek's API is significantly lower than that of other mainstream commercial large models and proprietary education models, lowering the entry barrier for education companies looking to develop AI-driven solutions.

With these advantages, DeepSeek has become a key driver of breakthroughs in AI-powered education. MoonFox Analysis selected Xueersi and Youdao as representative cases to further explore the new wave of AI education innovation driven by DeepSeek.

II. Xueersi: Taking DeepSeek as the Foundation, Integrating the Capabilities of the MathGPT to Achieve Dual Synergy and Ecosystem Development
Since February 8, Xueersi has successively integrated its AI-powered learning machines and learning & practice machines with DeepSeek, upgrading their AI capabilities through DeepSeek's deep thinking mode. Its flagship devices have already begun gray box testing, and free intelligent education features will gradually roll out. TAL also launched a new AI learning app, "Ask Me Anytime", jointly supported by its proprietary MathGPT and the DeepSeek large model, offering a one-stop intelligent education experience. Xueersi's AI education strategy on DeepSeek demonstrates dual synergy and ecosystem-oriented development in two key aspects:

1. Overcoming the limitations of a single model. By deeply integrating its education-specific large model with DeepSeek, Xueersi leverages DeepSeek's strengths in breaking down complex problems and enabling natural language interaction alongside the MathGPT's expertise in mathematical reasoning and subject knowledge mapping. This fusion allows for precise knowledge point analysis, positioning, and tracing; strengthened logical reasoning and manifested thought paths to understand and output multi-model content, so as to form the ability of heuristic guided thinking.
2. Hardware + software integration. Xueersi enhances traditional education hardware by using DeepSeek's deep thinking mode to address the common shortfall of focusing on answers rather than the thinking process. It also develops new AI learning applications that integrate self-developed instructional videos and frequently used AI learning tools, maximizing its accumulated educational content and large model technology advantages. Furthermore, Xueersi is exploring the application of DeepSeek in internal business scenarios, such as improving efficiency and reducing costs in customer service tutoring communications and video script creation.

By introducing DeepSeek for a multidimensional upgrade across "Hardware + Software + Content + Operations", Xueersi is poised to accelerate the iteration and enhancement of its product and service offerings, advancing the full AI-driven transformation of its K12 education business and further solidifying its leading position in the industry.

Xueersi's AI Education Strategy on DeepSeek

III. Youdao: Integrating Core Technologies to Build a Hybrid Architecture, Upgrading Its Entire Product Line, and Innovating AI-Native Learning Hardware
On February 6, Youdao announced the full integration of its product ecosystem with DeepSeek's inference capabilities, including Youdao XiaoP, Hi Echo, Youdao Dictionary, QAnything, and Youdao AI, enhancing the application of AI technology in educational settings. Later, on February 18, it launched the AI-native learning hardware product - the SpaceOne Dictionary Pen. By connecting its entire product line to DeepSeek and accelerating innovation in smart hardware, Youdao is driving the deepening of AI-powered education and optimizing operational costs.

1. Core Technology Integration: Youdao has built a hybrid architecture combining DeepSeek-R1's general-purpose intelligent inference capabilities with its self-developed education-specific large model "Ziyue". This integration enhances the precision of AI applications in complex educational scenarios and enables the development of multimodal applications based on DeepSeek.
2. Full Product Line Integration: On the software side, Youdao's AI learning apps have been integrated with DeepSeek, leveraging its deep thinking capabilities to improve AI-driven education and provide a more personalized interactive learning experience. Furthermore, Youdao's self-developed RAG engine "QAnything" based on Ziyue education large model and its AI open platform "Youdao AI" have fully adopted DeepSeek's inference capabilities for a comprehensive upgrade.
3. AI Education Hardware Innovation: Youdao introduced the AI-native learning hardware "SpaceOne" - a full-screen smart dictionary pen equipped with the DeepSeek-R1 inference model. It also comes preloaded with Youdao's AI tutoring software, educational resources, and a rich knowledge base, offering innovative in-depth explanations of challenging academic topics.

Youdao's AI Education Strategy on DeepSeek

Youdao's rapid and extensive adoption of DeepSeek spans across software (translation, speaking, and learning companion apps), hardware (dictionary pen), and AI platforms (Youdao AI), forming a comprehensive AI education ecosystem of "hardware as entry points + high-frequency software + foundational models". As DeepSeek integration deepens, Youdao is expected to accelerate the upgrade of its "Youdao XiaoP 2.0" and continue launching intelligent hardware products equipped with advanced inference large model capabilities.

IV. High-performance, Cost-effective General-purpose Models like DeepSeek to Drive AI Education Product and Scenario Innovation
The industry enthusiasm sparked by DeepSeek marks a critical turning point, where improvements in general-purpose large model capabilities and cost reductions enable the large-scale deployment of AI in education. Moving forward, general-purpose large models will empower AI education companies through two primary pathways:

• Path 1: Developing Proprietary Educational Large Models Enhanced by General-purpose Models like DeepSeek
  

By using general-purpose large language models like DeepSeek as the foundation and vertical educational large models as the core, companies can reduce issues like AI hallucinations in educational applications of general-purpose large models, ensuring professionalism and safety in AI-powered education. For instance, Squirrel Ai primarily relies on its self-developed vertical large model while integrating general-purpose large language models like DeepSeek, though the latter only accounts for about 10% of its product's AI capabilities.

• Path 2: Direct Collaboration with Foundational Large Models to Develop AI Education Products/Services
  

The path for education companies to directly adopt high-quality large models through cloud-based API calls for developing AI-driven education is becoming increasingly clear and feasible. By collaborating with leading large model providers, education companies can save costs and reduce staffing needs in areas like technology and computing power, allowing them to focus more on application scenarios. Leveraging their strengths in student learning data, educational operations, and industry insights, these companies can further explore innovative AI-driven education models and product services.

So far, DeepSeek's capabilities have been applied in smart educational hardware, AI education applications, educational content creation, customer and school-home communication, and internal business processes of educational companies. By combining general-purpose large model strengths with specialized educational data, DeepSeek overcomes application bottlenecks in certain subjects and educational scenarios, enabling cost reduction and efficiency gains. As the industry enters a new phase of deep AI integration, more native AI-powered educational software and hardware products are expected to emerge, transforming educational models and reshaping learning experiences. Smart educational hardware is likely to see breakthroughs first, making it a key area to focus on.

In terms of industry dynamics, against the backdrop of the rising popularity of the DeepSeek concept and widespread use of educational products, education companies that respond quickly to this wave of technological upgrades will attract more UVs, driving user growth and customer acquisition. In the medium to long term, DeepSeek empowers small and medium-sized educational institutions to deploy advanced AI with lower entry barriers, accelerating the industry's shift from a state of dominance by leading players to a more diversified competitive landscape. At the same time, the key to AI-driven education development will shift toward understanding educational scenarios and aligning models with business needs. In the future, the industry is expected to see the widespread adoption of hardware interfaces and the intelligent adaptation of educational content and services, reshaping the pace of innovation and the competitive landscape.

Primary Industry Preferences of DeepSeek Users in January 2025

Secondary Industry Preferences - Segments within Education & Learning

Data Source: MoonFox iApp
Data Cycle: January 2025

About MoonFox Data

As a sub-brand of Aurora Mobile, MoonFox Data is a leading expert in data insights and analysis services across all scenarios. With a comprehensive, stable, secure and compliant mobile big data foundation, as well as professional and precise data analysis technology and AI algorithms, MoonFox Data has launched iAPP, iBrand, iMarketing, Alternative Data and professional research and consulting services of MoonFox Research, aiming to help companies gain insights into market growth and make accurate business decisions.

About Aurora Mobile

Aurora Mobile (NASDAQ: JG) established in 2011, is a leading customer engagement and marketing technology service provider in China. Its business includes notification services, marketing growth, development tools, and data products.

For Media Inquiries:

Contact: zhouxt@jiguang.cn  | Website: http://www.moonfox.cn/en

Photos accompanying this announcement are available at

https://www.globenewswire.com/NewsRoom/AttachmentNg/7238a110-ef71-4fab-b510-816bf21fa9f3

https://www.globenewswire.com/NewsRoom/AttachmentNg/6958ca91-3df4-4bbf-b7d1-0e3a520136bd