(Yicai) Feb. 13 -- DeepSeek, the low-cost Chinese large language model described as an “AI Sputnik moment,” offers cost-effective artificial intelligence solutions for China’s small and medium-sized banks, which could help them to narrow the technology gap with larger rivals, according to banking industry insiders.

Among those that have embraced DeepSeek-R1, the latest iteration of the open-source AI model released on Jan. 20, are Bank of Beijing, Bank of Chongqing, Bank of Jiangsu, Chongqing Rural Commercial Bank, Postal Savings Bank of China, and Suzhou Merchants Bank. Except for PSBC, they are all mid-sized lenders.

DeepSeek-R1’s performance compares favourably with leading LLMs, but at much lower cost. Its application programming interface services are priced at CNY1 (14 US cents) per million tokens of input and CNY16 (USD2.20) per million tokens of output, much lower than OpenAI o1's rates of USD15 and USD60, respectively.

Due to the substantial resources needed to develop proprietary LLMs, small and medium-sized banks face the possibility of a widening technology divide with large banks. Integrating an existing low-cost LLM such as DeepSeek-R1 into their operations could give them the chance to narrow the gap.

DeepSeek can provide cost-effective smart business solution for smaller banks, enabling quick deployment without big investments in technological research and development, said Zeng Gang, director and chief expert at the Shanghai Institution for Finance and Development.

The AI can be rapidly deployed for use in areas such as intelligent risk control, contract verification, and customer insights, Zeng noted, slashing the time from introducing the technology to its actual practical use.

“DeepSeek's LLM technology has strong reasoning capabilities, efficient computational performance, and relatively low inference costs, making it well-suited for high-frequency calls and practical applications in specific scenarios,” a source at a city-level commercial bank told Yicai.

“This provides strong support for the application and development of AI in small- and medium-sized lenders,” the person noted.

A representative for Suzhou Merchants Bank said its intelligent decision-making system, built using the DeepSeek model series, has been used in more than 20 business scenarios, including credit risk control and anti-fraud monitoring. It has made the writing of due diligence reports 40 percent more efficient, the person added.

DeepSeek’s use in the banking sector is still in its infancy. The LLM's current uses are mainly focused on internal work scenarios, such as intelligent code writing, AI office work, and middle-office operations management, Yicai learned.

While these uses have enhanced employee work efficiency, they have not yet been applied to core businesses, including account transactions. Due to the extremely high requirements for data accuracy, the need to ensure model interpretability, and strict risk control, financial institutions tend to be cautious when applying LLMs, another financial industry source said.

Editor: Futura Costaglione