MoonFox Analysis | Large Models Implemented in K12 Online Education: Continuous Innovation and Upgrade of Product Functions in Off-campus Scenarios
Press Releases
Dec 17, 2024
About us: 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. As its sub-brand, MoonFox Data is a leading expert in data insights and analysis services across all scenarios, aiming to help companies gain market insights and empower precise decision-making.
SHENZHEN, China, Dec. 17, 2024 /PRNewswire/ —
I. Under the “Double Reduction” Policy, AIGC Applications Breathe New Life into the Education Industry
Since the implementation of the “Double Reduction” policy—which aims to reduce homework burdens and alleviate extracurricular tutoring pressures for students in compulsory education—the K-12 off-campus tutoring industry has gradually become more standardized. This policy emphasizes that off-campus training should serve as a beneficial supplement to school education, with “regulated and orderly” and “inclusive and high-quality” becoming the main themes for educational enterprises in their post-policy transformation.
With the advent of ChatGPT, “AI + Education” has emerged as a significant direction for the transformation of educational enterprises. In March 2024, the Ministry of Education issued four action plans to promote the integration of artificial intelligence in education. A key initiative is the “Implementation of Large AI Model Application Demonstration Action,” which encourages the use of large models in both classrooms and practical applications.
In July 2024, the Minister of Education announced at a press conference that the government would vigorously promote the construction of smart campuses, develop a Chinese version of the AI education large model, and explore large-scale personalized and innovative teaching methods to better meet public demand for “quality education.” In October 2024, the Beijing Municipal Education Commission, along with three other departments, released the Beijing AI Application Work Plan in the Education Sector, aiming to establish 100 benchmark schools for AI applications by 2025 and create several large models for the AI education industry by 2027. This reflects a strong push at both national and local levels for the implementation of large AI models in education.
The emergence of these large models has revitalized efforts to improve the quality of education in both school and off-campus settings, with a gradual shift toward practical applications. According to data from 21st Century Business Herald, by the end of June this year, approximately 40 large models and deep synthesis algorithms had been registered in the education sector. Leading companies like Baidu, iFLYTEK, and Zuoyebang have already integrated large models into their online education applications, making these innovations a core competitive advantage in the AIGC era of online education. Aurora Research Institute has summarized the high-frequency functions of AI large models in educational applications, categorized by main scenarios and user groups, as follows:
Function Innovations Brought by Large Models
On-campus |
Off-campus |
||
Student |
Teacher |
Student |
Parents |
AI Error Book |
Intelligent Question Generation |
Question & Answer Search |
Intelligent Paper Correction |
Learning Situation Analysis & Ability Diagnosis |
Intelligent Paper Correction & Examination |
AI Oral Practice Partner |
Learning Situation Analysis |
AR Teacher/Virtual Classroom |
Intelligent Teaching Design |
AI Learning Companion |
|
Intelligent Teaching Resource Management |
AI Teacher |
||
Intelligent Teaching Assistance |
Open Question & Answer Dialogue |
||
Classroom Situation Analysis |
AI Writing Assistant |
II. Growth in K-12 Off-Campus Online Education Usage and Market Development: Insights from the Top 5 Applications
From the application perspective, the current K-12 education market is primarily driven by off-campus online education applications (excluding home-school interaction apps), which have demonstrated periodic growth in both monthly usage time and launch frequency throughout 2024.
Monthly Usage Time for K12 Education Applications (Unit: 10,000 Hours)
Date |
Off-campus |
Others |
2024.01 |
27,251.11 |
2,779.72 |
2024.02 |
24,194.87 |
4,088.36 |
2024.03 |
24,180.36 |
2,510 |
2024.04 |
24,317.76 |
2,313.95 |
2024.05 |
24,015.19 |
2,395.51 |
2024.06 |
24,320.55 |
2,722.42 |
2024.07 |
20,020.61 |
6,297.45 |
2024.08 |
19,295.09 |
5,830.8 |
2024.09 |
21,626.15 |
2,692.04 |
2024.10 |
27,183 |
3,306.7 |
Data Source: MoonFox iApp; Off-campus online education applications exclude home-school interaction apps. |
Data Cycle: January 2024 – October 2024 |
Monthly Launch Frequency for K12 Education Applications (Unit: 10,000 times)
Month |
Off-campus |
Others |
2024.01 |
230,236.36 |
24,629.25 |
2024.02 |
208,599.13 |
24,866.12 |
2024.03 |
222,569.19 |
24,292.94 |
2024.04 |
216,062.89 |
24,071.93 |
2024.05 |
230,876.12 |
24,697.32 |
2024.06 |
226,144.4 |
26,314.17 |
2024.07 |
193,056.73 |
32,928.59 |
2024.08 |
176,918.18 |
26,096.02 |
2024.09 |
210,321.03 |
24,281.2 |
2024.10 |
256,775.41 |
27,022.78 |
Data Source: MoonFox iApp; Off-campus online education applications exclude home-school interaction apps. |
Data Cycle: January 2024 – October 2024 |
Analyzing data from the top five off-campus education apps, we find that their market penetration in K-12 online education has reached two-thirds. The monthly active user numbers for these apps remain relatively stable, with notable increases in new users following the start of the new school year. This suggests that the off-campus education application market still has considerable growth potential. The user base for these top five applications is primarily concentrated in markets outside Tier 1 and emerging Tier 1 cities, aligning with their positioning to address the unequal distribution of educational resources.
Monthly New User Comparison for Top 5 Off-Campus Education Apps (Unit: Individuals)
Date |
Zuoyebang |
Xiaoyuan Oral Calculation |
Onion Academy |
Xiaoyuan Q&A Search |
Zuoyebang Parent Version |
2023.11 |
8,543,948 |
1,473,617 |
950,479 |
1,066,104 |
1,241,996 |
2023.12 |
9,724,221 |
1,738,452 |
893,384 |
1,041,686 |
1,059,410 |
2024.01 |
9,780,960 |
1,237,051 |
1,341,642 |
1,014,617 |
605,876 |
2024.02 |
9,517,768 |
1,316,957 |
1,122,342 |
867,725 |
589,302 |
2024.03 |
10,339,111 |
1,645,887 |
1,185,884 |
914,823 |
818,512 |
2024.04 |
9,756,630 |
1,608,489 |
1,343,767 |
864,681 |
1,637,942 |
2024.05 |
8,808,864 |
2,087,424 |
1,156,881 |
885,782 |
2,952,580 |
2024.06 |
8,478,147 |
2,903,895 |
1,403,468 |
1,096,040 |
2,936,008 |
2024.07 |
7,286,435 |
1,584,647 |
3,155,486 |
980,370 |
743,095 |
2024.08 |
7,051,982 |
1,685,746 |
2,080,305 |
821,504 |
670,791 |
2024.09 |
13,412,940 |
3,695,581 |
1,748,740 |
1,163,060 |
771,313 |
2024.10 |
15,871,114 |
9,783,197 |
1,337,583 |
1,757,086 |
995,938 |
Data Source: MoonFox iApp; |
Data Cycle: November 2023 – October 2024 |
City Level Distribution of Top 5 Off-Campus Online Education Applications
City Level |
Zuoyebang |
Xiaoyuan Oral Calculation |
Onion Academy |
Xiaoyuan Q&A Search |
Zuoyebang Parent Version |
First-tier Cities |
4.617 % |
6.095 % |
5.934 % |
5.691 % |
4.326 % |
New First-tier Cities |
15.726 % |
20.789 % |
14.899 % |
18.346 % |
15.326 % |
Second-tier Cities |
18.562 % |
23.148 % |
18.242 % |
20.267 % |
17.165 % |
Third-tier Cities |
26.684 % |
21.681 % |
26.922 % |
24.282 % |
27.318 % |
Fourth-tier Cities |
21.627 % |
18.105 % |
21.351 % |
19.944 % |
22.253 % |
Others |
12.784 % |
10.182 % |
12.653 % |
11.470 % |
13.612 % |
Data Source: MoonFox iApp; |
Data Cycle: October 2024 |
The top five applications have adopted a “combination strategy” to expand their market share. They have developed a comprehensive solution that integrates “software, hardware, curriculum systems, and large models.” This approach begins with existing educational app services and enhances them by incorporating hardware solutions based on current language large models. As a result, this strategy not only improves the product experience for existing users but also attracts new users through various appealing factors.
Ecosystem Layout of Online Education Enterprises behind the Top 5 Applications
Application |
Hardware |
Hardware Shipments |
Software |
Large Models |
Zuoyebang Zuoyebang Parent Version |
Learning machines, tutoring machines, learning pens, learning desks, word cards, learning printers |
Over 2 million units of educational smart hardware shipped in 2022 |
Software ecosystem covering both academic and non-academic subjects |
Galaxy Large Model |
Onion Academy |
Onion Planet learning machine |
200,000 units sold |
Intelligent educational services, AI learning companion services |
Doubao Large Model |
Xiaoyuan Oral Calculation Xiaoyuan Q&A Search |
Xiaoyuan learning laptop |
Over 1 million units |
Focused on online course systems, both academic and non-academic software ecosystem |
Feixiang Large Model |
III. Large Model Implementation in Off-Campus Online Education: Driving Functional Upgrades and Innovations with “Inspiration” and “Guidance” as Core Concepts
From the perspective of leading off-campus online education solutions, the primary growth drivers in the industry are the functional upgrades and innovations enabled by “applications + large models.” These innovations focus on several key areas, including Question & Answer Search, AI Oral Practice Partners, AI Open Question & Answer Dialogue, AI Learning Companions, AI Teachers, AI Writing Assistants, Intelligent Paper Correction, and Learning Situation Analysis.
Question & Answer Search:
Current solutions have evolved beyond merely searching for answers from a vast question bank. They now leverage large models to analyze and reason, extracting key information from questions and accurately understanding the relevant knowledge points. This enables them to search for answers using a Socratic-style approach, which aims to inspire and guide students while cultivating proper thinking patterns and habits.
After providing answers, these solutions offer personalized recommendations for similar questions based on knowledge points and question types. Both online education applications and AIGC solutions have enhanced their question-and-answer search capabilities. For example, online education apps like Zuoyebang and Yuanfudao utilize extensive question banks to offer photo-based Q&A search services, which have further developed into one-on-one tutoring, VIP video lessons, and live streaming courses to enrich their curriculum. Meanwhile, AIGC applications focus on multimodal input (voice, text, images), reasoning and guidance for new or challenging questions, and emotionally nuanced communication, as seen in products like Xueersi MathGPT, Doubao Learning, Quark AI Q&A Search, and ERNIE Large Model.
AI Oral Practice Partner:
AI technology simulates real-world conversational scenarios by providing multiple conversation settings. Leveraging knowledge graph technology, it connects various knowledge points, offering a diverse range of topics for students. With speech recognition and grammar analysis capabilities, it accurately detects pronunciation and grammatical errors, providing relevant suggestions, as well as tips for oral expression and common phrases. Additionally, it features sentence repetition functionalities with “guidance + voice prompts.”Key players in the AI Oral Practice Partner space include English education apps like Youdao’s Ziyue Large Model, search engines such as Baidu and Google, and AIGC assistants like Tencent Yuanbao.
Open Dialogue:
Open dialogue primarily relies on large models, natural language processing, and knowledge graph technologies. It utilizes big data from student behaviors to facilitate exploratory conversations, such as AI celebrity dialogues, which guide students to think critically while imparting knowledge. This approach fosters proactive questioning, thinking, and exploration skills. Currently, Yuanfudao’s “Dolphin AI Learning” offers features like AI celebrity “cross-time and space dialogues.”
AI Learning Companion:
The AI learning companion feature is one of the most innovative advancements, combining the capabilities of large models to provide students with personalized, emotional, and comprehensive educational support. Tailored learning strategies are offered based on students’ learning abilities, progress, and behavioral data. These strategies include online homework grading, error correction, intelligent analysis of learning performance, and customized training, covering various scenarios such as daily practice and exam preparation. Some solutions also incorporate features like task timing and background music, which guide students in developing structured learning paths and cultivating focused study habits. Additionally, these companions can support emotional communication and provide personalized counseling for adolescent mental health. Current mature AI learning companion solutions include Onion Academy, Xueersi’s “Xiao Si,” and Sharing School’s “Xiaoshu Learning Companion.”
AI Teacher:
The emergence of AI teachers is transforming educational applications from a cramming-style teaching model to an interactive teaching approach. By designing questions and facilitating communication, AI teachers can accurately identify students’ weaknesses and provide real-time board writing and video explanations. This helps students efficiently complete key learning tasks such as knowledge previews, homework checks, exam reviews, and test preparation. One notable example is the “Super Human One-on-One AI Teacher” developed by Jingzhunxue.
AI Writing Assistant:
AI writing assistants can be categorized into three main functions: AI writing guidance, AI writing correction, and AI sample generation. AI serves as a writing guide by helping students analyze topics, determine main ideas, and organize their writing thoughts, much like the writing assistance feature in QQ Browser.
AI correction, which previously relied on OCR (Optical Character Recognition) technology for basic oral math corrections or error judgments in application-generated questions, has evolved into a robust system that provides guidance and corrections for both Chinese and English essays using large models. This AI writing correction incorporates expert teaching methods to assess content, offering detailed feedback and suggestions to enhance writing skills. For example, Yuanfudao’s AI writing correction evaluates various aspects of an essay, including topic analysis, idea formulation, sentence structure, and vocabulary. Additionally, Youdao Writing is an intelligent English writing correction tool that can identify over 100 types of errors, including phrase collocations, tenses, and subject-verb-object agreement. AIGC solutions are particularly effective in generating model essays.
Learning Situation Analysis:
Learning situation analysis generates real-time records of students’ extracurricular activities, analyzing data such as error statistics, reasons for mistakes, mastery of knowledge points, and the development of skills and competencies. It regularly produces highly readable learning reports and feedback for parents. Most AI education applications now incorporate learning situation analysis functionality.
In summary, large models serve as a chain connecting individual functionalities, supporting multimodal input and delivering guiding, companion-like outputs. Q&A search capabilities extend beyond traditional question banks, answers go beyond simple step-by-step outputs, and dialogues are not restricted to fixed topics and contexts. AI capabilities encourage students to develop independent thinking, driving the transformation of educational tutoring from merely intelligent to genuinely wise. Online education companies with extensive experience in educational applications, along with large model technology firms specializing in search engines and voice interaction, will be the main players in this space. Their competition and collaboration will further propel the intelligent development of the online education industry.
- Our Information:
Website: https://www.moonfox.cn/
Contact number: 400-888-0936
- Contact us:
Name: Felix
Title: Director of Sales, Industry Insight Division
Tel: +86 -13366276383
Email: zhouzt@jiguang.cn
Address: 608/F, Tower B, Wintrust Center, No. 1 Xidawang Road, Chaoyang District, Beijing, China
Name: Janette Zhou
Title: Marketing Manager
Tel: +86-13928239332
Email: zhouxt@jiguang,cn
Address: 608/F, Tower B, Wintrust Center, No. 1 Xidawang Road, Chaoyang District, Beijing, China
SOURCE Aurora Mobile Ltd