Reported by 2 outlets
The story so far
AI summary of the reporting below · Updated · How we work
Designed for Nvidia B200 GPUs, the recipes coordinate request routing and multi-GPU communications across data center clusters, which Nvidia claims can boost out-of-the-box performance by up to 2x. [1][2]
StorageReview noted that Nvidia claimed the 2x performance figure without detailing baseline metrics, workloads, or serving setups. Alongside this release, LG AI Research is offering API access to K-EXAONE-2.5, which used Nvidia's Nemotron-Personas-Korea dataset of 7 million synthetic personas. [1]
The software release accompanied wider regional partnerships with South Korea's Ministry of Science and ICT covering agent security and sovereign infrastructure. Hyundai Motor Group is evaluating Nvidia's OpenShell runtime and NemoClaw blueprints for enterprise agents, while Korean security firm AIM Intelligence is connecting its Starfort platform with OpenShell. In infrastructure, a planned collaboration between Rebellions and SK Telecom targets 2028 systems pairing Nvidia GPUs for prefill with Rebellions NPUs for decode. [1]
NVIDIA is releasing Dynamo-optimized deployment recipes for Korea’s leading 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. As models grow beyond the capacity of a single GPU, they must be split across multiple GPUs, requiring careful coordination of request routing…
via NVIDIA
NVIDIA released Dynamo deployment recipes for four Korean open models at AI Day Seoul: SK Telecom’s A.X K2, LG AI Research’s K-EXAONE 2.0, Upstage’s Solar Open 2, and Motif Technologies’ Motif-3. The recipes are tuned for NVIDIA B200 GPUs, and NVIDIA says they deliver up to 2x performance out of the box without naming the
via StorageReview