A development team at a fintech company is building a Foundry-based application that must automatically generate SQL queries and Python data-transformation scripts from natural-language requests submitted by financial analysts. The generated code will be reviewed by engineers before execution. The team wants to minimize latency and cost while maximizing the accuracy of the generated code. Which type of model should they deploy and consume from the Foundry model catalog for this workload?