The wave brought by DeepSeek has spread to public funds.
According to reports, as of now, more than ten public funds, including Fidelity Fund, Bosera Fund, Guotai Fund, Tianhong Fund, Huatai-Pb Fund, China Europe Fund, Yongying Fund, Wanjia Fund, Invesco Great Wall Fund, and Nuoan Fund, have carried out the privatization deployment of the DeepSeek series of open-source models.
By the way, for friends who want to seize this wave of opportunities to make money, I recommend using biyapay for transactions, as biyapay does not freeze cards for deposits and withdrawals.
Another industry insider stated, ‘The company has been working early in the AI direction and has already ventured into large models. There may also be privatization deployment of the DeepSeek series of open-source models in the future, but we are still observing.’ Some small and medium-sized public fund personnel pointed out that investment in computing power inevitably encounters high cost expenditure issues, which is also a difficulty, but if one wants to improve business efficiency and achieve digital transformation, attempts to layout in the technology field are essential.
However, several foreign public funds interviewed stated that they have no deployment plans for now. Another institutional insider, somewhat mysteriously, stated, ‘The company has applications for AI large models deployed globally, but typically does not disclose which specific models are chosen.’
Mainly applied to core business scenarios such as investment research, consulting services, and product sales.
From the perspective of application scenarios, the privatization deployment of the DeepSeek series of open-source models by public funds mainly includes core business scenarios such as investment research, consulting services, software development, risk management, document processing, and product sales.
Bosera Fund stated that as early as the beginning of 2024, it discovered the potential of the DeepSeek model in automatically writing code and logical reasoning, and was the first to deploy the DeepSeek-v1 model on its own Ascend server as the base model for the company's smart development tools, which was later upgraded to the DeepSeek-v2 model in August 2024.
At the beginning of 2025, with the release of the DeepSeek-R1 model, Bosera Fund completed internal deployment and began exploring its applications in investment research, consulting services, and software development.
Bosera Fund believes that the R1 model performs exceptionally well in reasoning capabilities, which can help enhance work efficiency and support business innovation. At the same time, its demand for computing power has also decreased, creating conditions for further promotion and application.
Guotai Fund stated that by the end of January 2025, the company completed the privatization deployment of the DeepSeek series models and established the Guotai Fund AI application development platform, achieving impressive results in business scenarios such as brand maintenance, risk management, product operations, and document processing. As a result, business personnel can easily and quickly conduct in-depth analysis of relevant reports, strengthen data mining capabilities, streamline information flows, and improve operational efficiency.
Huatai-Pb Fund also recently announced that it has applied the DeepSeek series of open-source models to core business scenarios such as investment research, product sales, risk control compliance, and customer service. From intelligent assistance in investment research to personalized experience enhancement in customer service and precise management in risk control, large model technology is deeply integrated into various business segments of the company.
Taking e-commerce as an example, Huatai-Pb Fund's internet finance department has formed an AI team codenamed 'deepfund,' utilizing AI large models to comprehensively enhance the operational efficiency and user experience of fund e-commerce, promoting the intelligent development of e-commerce sales. The company is also conducting testing and validation of large model applications across multiple business scenarios, ensuring the stability and reliability of large model applications through rigorous testing processes and data analysis, laying a solid foundation for subsequent large-scale promotion and application.
Fidelity Fund stated that its technology team has actively explored the application of large language models and deployed multiple open-source models, including DeepSeek. After exploration and verification, localized deployment models have reached a usable stage in application directions such as internal data processing, code assistance generation, text generation, enterprise-level RAG, and research report interpretation. With the iteration of models and further integration of AI applications, existing workflow experiences will be comprehensively optimized, enhancing work efficiency.
Tianhong Fund is currently conducting comprehensive follow-ups on DeepSeek's related technologies, such as testing the business effects of its models, utilizing its training for Tianhong Fund's own large models, and developing reinforcement learning paths, with some exploratory progress already made.
After the release of DeepSeek R1, Wanjia Fund quickly completed the integration of the official website API and successfully deployed the 32B local model. Through its self-developed platform 'WanChat,' which has multi-model dialogue capabilities, it can quickly compare the actual performance of different models.
After preliminary testing by Wanjia Fund's marketing department, the model has demonstrated outstanding performance in the field of Chinese text processing, especially in scenarios such as summarization and document generation, surpassing international benchmarks like ChatGPT and showcasing breakthrough advancements in localized AI technology. Currently, Wanjia Fund is empowering core business segments such as risk management and wealth management through a three-stage strategic transition of 'informatization - digitalization - intelligentization.'
Nuoan Fund has built an AI capability platform covering functions such as image content extraction, speech-to-text, text-to-speech, and document summarization through the privatization deployment of small models. At the same time, it has independently developed a unified AI model gateway based on open-source frameworks to achieve efficient management of multi-model collaborative scheduling and AI resources.
‘The impact of the DeepSeek series of open-source models is disruptive.’
After deeply embracing DeepSeek, will AI disrupt the landscape of the fund industry?
An insider from a public fund in Shanghai pointed out that the impact of the DeepSeek series of open-source models is disruptive. If utilized properly, it can significantly reduce costs in public systems, personnel, and more, such as using AI for posters, videos, etc. ‘It may be a bit exaggerated to say that the privatization deployment of the DeepSeek series of open-source models can reshape the public fund landscape, but there may still be a gap in the future.’
‘In fact, when handling work now, I always prioritize searching and generating on the DeepSeek software. This is also a requirement I set for myself: to use AI as much as possible in my work, even if it means forced use,’ said an insider from a certain public fund in South China.
The above-mentioned person further stated that the application scenarios of AI in the asset management field are also extremely broad. For example, in investment research, under the traditional business model, researchers and fund managers need to process vast amounts of information to transform fundamental information into investment decision references. However, with the help of AI tools, the efficiency of this process has been significantly improved. In terms of risk control, financial institutions are essentially more like technology companies. With the development of financial technology to this point, AI has comprehensive value in preventing risks, particularly in deep learning and optimizing the risk-return ratio of investment portfolios. Furthermore, in operations, especially in the content production field of marketing, DeepSeek can actually achieve some degree of complete replacement.
‘I don't know if the emergence of DeepSeek is that definitive node, but there is no doubt that we are in an era of transformation, and AI is the core force driving change. This trend is very clear: first, the iteration of algorithms and the enhancement of computing power support the availability of applications, including data processing functions needed by the financial industry, which AI excels at; second, industry competition is intensifying, and the role of AI in cost reduction and efficiency enhancement is indisputable; finally, the upgrading of customer demands and the challenges faced by industry development also require a technological revolution like AI to compel upgrades and iterations.’ At the same time, the person also emphasized, ‘We will not be replaced by AI, but we will be replaced by those who learn to use AI first.’
From the current application effects, Yongying Fund analyzed that DeepSeek performs excellently in aspects such as thinking process, Q&A effectiveness, and response speed. For example, in terms of Q&A effectiveness, DeepSeek's accuracy rate has reached over 90%, significantly higher than that of other similar models. At the same time, its deep thinking ability simplifies the difficulty of inputting prompts, making it easy for non-professionals to use. Additionally, models like 7B and 14B have significantly reduced requirements for computing power, lowering deployment costs.
However, Yongying Fund also pointed out that the current DeepSeek models mainly focus on text analysis capabilities, and there are still deficiencies in multimodal functions (such as image recognition and speech processing). Future research will focus on optimizing these functions to enhance the overall performance of the models.
‘In the process of introducing models, it is essential to build a stable and efficient foundational technology platform and talent team, while also validating the application of AI in business scenarios, which is a prerequisite for AI application exploration across industries,’ said Fidelity Fund.
At the same time, Guotai Fund emphasized that with the rapid development of science and technology and the widespread application of artificial intelligence, the company will continue to pay attention to its information security and compliance risks, adhering to safety as a principle in the integration and control of technology and business, strictly auditing potential risks, and implementing multiple technical safety lines to effectively safeguard the first line of defense for fund information security.