NVIDIA Releases Dynamo-Optimized Deployment Recipes for Four Korean Open Models

NVIDIA is releasing Dynamo-optimized deployment recipes for four Korean open models: A.X K2 from SK Telecom, K-EXAONE 2.0 from LG AI Research, Solar Open 2 from Upstage and Motif-3 from Motif Technologies. The recipes are designed for multi-GPU and multi-node inference, where models must be split across GPUs and routed carefully to preserve performance. Optimized for NVIDIA B200 GPUs, the company says the recipes can deliver up to 2x performance out of the box and help developers move these models into production faster.