Financial institutions are prioritizing open-source gen AI models for operations. 

Open-source AI models are public, making them more customizable compared with closed-source models, in which only the creator can make appreciable changes. Open-source models also can run locally without external connection. 

Eighty-four percent of FIs prefer open-source models, according to the “AI in financial services 2026” report by Nvidia, published today. The chip maker surveyed 839 leaders in financial services from retail banking, asset management, insurance and insurance between August and September 2025.

“This has probably been one of the bigger pendulum swings that we’ve seen in the financial services industry since ChatGPT came on the scene,” Kevin Levitt, global business development lead for financial services at Nvidia, told FinAi News. 

“Three years ago, a lot of companies rushed to get generative AI solutions to market it to their leadership and board,” he said.  

Now FIs are more selective about what kind of models they want to deploy, he said. 

When pairing an open-source model with a bank’s proprietary data, outcomes and results are more accurate, Levitt said. This increases  

• Productivity;  

• Security;  

• Revenue gains;  

• Efficiencies; and  

• How the bank operates. 

Expense management 

FIs prefer open-source models for expense management, Levitt said.  

In fact, they will pay “inordinately more” when using an off-the-shelf solution by a third-party vendor, compared with an open-source model which they own and can customize, he said. 

To meet changing industry demands, Nvidia has developed its Inference Micro-Service solution, which runs at the bank and provides higher security controls than closed-source models, Levitt said. 

Open-source AI models include: 

• France-based MistralAI;  

• Chinese model DeepSeek; 

• Alibaba’s Qwen; and  

• Meta’s Llama. 

Increasing demand for AI factories 

While AI model preferences change, the plan to deploy AI is ever-growing. 

Nearly 100% of FIs plan to increase or keep their AI spend constant in 2026, according to Nvidia’s report. 

Many financial institutions are looking for end-to-end AI stacks that boost productivity and expense management, Levitt said, adding that Nvidia’s AI factories have seen a steady increase in demand from the industry. 

FIs seeing ROI from AI investments 

FIs are also seeing increased return on investment from AI, which incentivizes them to invest more in the technology, Levitt said. 

In fact, 89% of survey respondents said AI has helped increase annual revenue or decrease annual costs. 

Customer service, algorithmic trading, document processes and payments operations are among the biggest areas where ROI is observable, according to the report. 

Register here for the inaugural FinAi Banking Summit, taking place March 2-3 in Denver. View the full event agenda here.