The year 2024 is widely regarded as a significant turning point for the financial industry. In this year, the internet technology marked a thirty-year milestone, and at the same time, the field of financial technology witnessed the accumulation of its first decade of achievements. The rising trend of large model technology, the innovative engine of fintech, is drawing a new yet challenging blueprint for the industry's development.

As part of the historical process of technological innovation, financial enterprises are embracing the era of large models with unprecedented speed and posture. However, this transformation is not without its obstacles. The approaching challenges include: how to continuously improve computing capabilities under the condition of limited computational resources? When making choices between open-source and commercial models, and seeking a balance between general and specialized large models, what should be the considerations? How to ensure the efficient operation of Retrieval-Augmented Generation (RAG) systems in terms of real-time response and meticulousness? Furthermore, how can the security and compliance of large models be fully guaranteed? These are just a few of the questions that financial enterprises must answer on the new era's path.

At such a historic juncture, Alibaba Cloud released "Refined by Fire: A New Chapter for Large Financial Models", providing the financial technology industry with a series of thought and action directions. The report deeply analyzes how financial enterprises can successfully embrace large models, explicates the developmental trends inherent in large models, and showcases the blueprint for constructing a core technology landscape of financial-grade AI.

The term "Refined by Fire" in the report suggests the constant refinement and improvement of technology, while "Turning to Gold" symbolizes the value creation and qualitative leaps brought about by the application of large models. Highlights of the report include an analysis of five distinctive characteristics of the development trend of large models:

• First, the "Cloud+AI" architecture highlights the increasingly close integration of large models with cloud computing. This combination, particularly in tech innovation hubs like the United States and China, has significantly facilitated the training and rapid deployment of large models, accelerating the innovation cycle of AI.
• Second, the "AI Everywhere" concept means that large models are becoming omnipresent, forming a diverse ecosystem with small models, new types of endpoints, and data platforms, deeply integrating into everyday business processes and user experiences.
• Third, the enterprise market's favoring of AI Native SaaS shows that large models have become an important force driving the enterprise market towards specialization and vertical development.
• Fourth, the rapid enhancement of the functionalities of large models, with the application of multimodal capabilities, Agent modes, and Assistant APIs, becoming the core of technological breakthroughs.
• Fifth, privacy and data security have gained more attention in the field of large models, establishing a comprehensive security assessment system along with data classification and grading.

Overall, these characteristics not only provide guidance for the future development direction of AI technology and applications but also clearly indicate the direction for the intelligent upgrading of financial enterprises.

Financial institutions are the core support of the national economy. Their operating efficiency, risk management, and service quality play a crucial role in the stability and growth of the economic and social environment. Faced with the emerging trend of "large model" technologies, building and enhancing artificial intelligence (AI) systems and applications that can both meet the stringent demands of the financial industry and possess cutting-edge technology levels has become an urgent subject. This involves not only the high development of the technology itself but also demands that safety, reliability, scalability, and compliance must meet the stringent standards of the financial industry.

In response to the complex demands of financial services, the report released by Alibaba Cloud deeply analyses the core challenges currently faced by financial customers, puts forward strategic planning for the construction of financial-grade AI, and six key elements. In addition, the report also identifies three application scenarios with significant commercial value: knowledge-intensive business, full-process digitalization, and multimodal digitalization. It also provides a detailed introduction to the technical architecture behind these scenarios and their implementation paths, aimed at accelerating the pace of digital and intelligent innovation for financial enterprises.

It is expected that by 2024, the era of large models in the financial industry will officially begin. Although facing many challenges, with clear strategic guidance and solid technical support, financial enterprises will gradually enter a new phase of intelligent upgrading and move towards a future of "integrated cloud data intelligence."