Nota supplies optimized LLM, servers for Dube financial coding agent
Nota said delivery is complete and that the same GPU setup now handles up to 95% more requests per minute for Dube's on-premises agent.
Nota has signed a contract with Dube to supply an optimized large language model and servers for an on-premises AI coding agent built for the financial sector, and said delivery under the contract has been completed. The deal was reported by datanet.co.kr.
Under the contract, Nota provided model lightweighting, inference and serving environment optimization, and server configuration for Dube's AI coding agent. Dube is a financial-sector SI development company that supplies AI coding agents to financial institutions.
Nota said requests per minute rose about 58% to 95% on the same GPU configuration compared with before optimization, and that first-token latency was cut to less than half.
The company used quantization to reduce the model's memory requirements and optimized both the inference process and the serving environment.
Quick answers
What did Nota supply to Dube?
An optimized LLM and servers for Dube's on-premises AI coding agent for the financial sector, including model lightweighting, inference and serving optimization, and server configuration.
How much did performance improve?
Nota said requests per minute rose about 58% to 95% on the same GPU configuration, and first-token latency fell to less than half.