Yunqi Capital | Self-Developed Large Model Gets Regulatory Approval: IceKredit in the Post-AI Era
How to Embrace the New Wave?
OpenAI's service suspension rules have opened up new possibilities for the large model ecosystem, drawing more attention to the commercialization progress of domestic LLMs.
In mid-June, Origin One, the large model launched by Yunqi Capital portfolio company IceKredit, received regulatory approval for generative AI services. This followed its deep synthesis service algorithm filing in January, achieving dual compliance certification for both "algorithm" and "service". It also signals the beginning of accelerated large-scale industrialization and deep application for Origin One.
Origin One had already accumulated numerous overseas application cases in finance, e-commerce, and office productivity. In this edition of Yunqi Capital, we take you inside Origin One to explore how IceKredit — a company that has used AI to empower its work from Day 1 — is embracing the new wave of AI.
"IceKredit is essentially a technology platform that insists on using artificial intelligence as its algorithmic foundation." In 2020, before ChatGPT ignited the current AI boom, company founder Lingyun Gu emphasized this AI element in the company's technical base during a conversation with Yunqi Capital.
In 2023, when generative AI became the dominant theme of technological transformation, IceKredit — which had long focused on AI + finance — actively embraced the new wave. Building on years of accumulated expertise in machine learning, deep learning, and natural language processing, IceKredit launched its large language model Origin One, beginning to unlock new possibilities for providing enterprise-grade services using AI technology across multiple domains including finance.
Origin One serves both B2B and B2C functions, with core capabilities including multi-turn dialogue, knowledge Q&A, code generation, text summarization, creative writing, language translation, and data reasoning. One distinguishing feature is its integration of domain-specific Know-How from multiple vertical industries with large model capabilities, thereby empowering digital transformation across various sectors. Additionally, Origin One supports both API integration and private deployment, providing users with more stable, secure, and reliable service.
Finance: Empowering Document Processing, Knowledge Base Construction, and More
For AI products serving the financial industry, strict regulation and professionalism are unavoidable characteristics. This means that only by training on professional data with domain expertise can vertical models suitable for industry needs be developed.
IceKredit has years of algorithmic model experience and industry Know-How serving banking and insurance clients. Building on this foundation, the Origin One large model possesses capabilities in financial professional Q&A, logical reasoning, and ultra-long text processing, providing AI-powered support for financial document processing, analysis of foreign investment products, and other financial business scenarios.
At the same time, IceKredit has a complete standardized solution for "knowledge base construction + vertical large model training," which not only helps financial institutions consolidate and organize internal business knowledge bases, but also combines proprietary business knowledge to rapidly train exclusive vertical scenario large models, deeply satisfying customers' customized needs.
Customer Service: Alleviating Human Resource Gaps and Optimizing Processes
Origin One can capture long-term dependencies in text, understanding context and meaning to generate more accurate, coherent, and semantically rich text. It has already been deployed in multiple customer service scenarios.
For example, in the intelligent customer service system of an international e-commerce platform, the Origin One large model can help replace non-sensitive portions of customer service work, including pre-sales/mid-sales/after-sales consultation responses, order query responses, logistics queries, return and exchange processing, complaint handling, work order tagging, and customer service work review and analysis — providing real-time response and 24/7 support while alleviating gaps and pressure in human customer service staffing.
Origin One has also been successfully applied in international insurance business scenarios. Based on phone recordings provided by the client company, it rates customer service representatives' work performance, helping the company establish an evaluation system for representative performance and optimize their workflows.
Education: Blooming Across Translation, Writing, and Creative Work
Today, generative AI is shining in educational scenarios such as translation, writing, and assisted instruction, and Origin One's AI capabilities align with the needs of these niche applications in many ways.
For instance, it supports direct translation of files in multiple formats; it features a writing assistant function that can provide creative ideas and content based on user-input themes and writing requirements, and also supports grammar, vocabulary correction, and polishing for user-uploaded essays. Additionally, Origin One has launched a "Story Loop" feature that can provide unique perspectives, plotlines, and descriptive copy based on user-input themes and story backgrounds, assisting users in creating fairy tale texts.
Beyond these domains, Origin One's AI solutions are also solving problems for clients in marketing, office productivity, and other fields, with its service boundaries continuing to expand. We look forward to IceKredit leveraging its deep technical accumulation and vertical industry Know-How to validate more possibilities for artificial general intelligence.
· References
-
Is Intelligent Customer Service "Losing Its Mind"? Large Language Models Offer Solutions. IceKredit Research Institute. https://www.icekredit.com/research/1785
-
Generative AI Accelerates Penetration in Education Industry, Could Intelligent Learning Hardware Become a Commercialization Breakthrough? IceKredit Research Institute. https://www.icekredit.com/research/1797
