FICO today announced a new suite of domain-specific generative AI models designed to improve accuracy and trust in financial services applications.

The FICO Foundation Model for Financial Services (FFM) includes two offerings: the Focused Language Model (FLM) and the Focused Sequence Model (FSM), according to today’s FICO release. Both are built on curated, task-specific data sets to reduce hallucinations common in general-purpose large language models.

Unlike traditional LLMs, FICO’s models require significantly fewer computing resources — up to 1,000 times less — making them cost-effective, auditable and adaptable, the company said. The models also incorporate patented “Trust Scores,” allowing financial institutions to set risk thresholds and ensure regulatory compliance.

“The focused foundation model represents a practitioner’s approach to GenAI in financial services,” said Scott Zoldi, chief analytics officer at FICO. “These models provide transparency, auditability and adaptability while mitigating hallucinations.”

FLM is designed to enhance compliance, fraud detection and customer interactions, while FSM analyzes transaction sequences to uncover patterns in areas such as payments and risk assessment. FICO said the models have demonstrated compliance improvements of up to 38% and accuracy lifts of more than 35% in transaction analytics.

Check out our exclusive new bank industry data here.