NVIDIA has published deployment configurations for four open models developed in South Korea, giving developers a way to run inference across multiple B200 GPUs when a model outgrows one processor.
The models are SK Telecom’s A.X K2, LG AI Research’s K-EXAONE 2.0, Upstage’s Solar Open 2 and Motif Technologies’ Motif-3.
NVIDIA’s open-source Dynamo framework routes requests among GPU workers and coordinates inference across nodes.
The recipes add model-specific deployment guidance for Dynamo. On Nov. 5, Supermicro, IBM and NVIDIA announced a shared key-value cache system using Dynamo, aimed at reusing prompt data and reducing processing time for long-context AI inference.
