Nvidia released 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, according to a company blog post.
Nvidia Dynamo is an open-source distributed inference-serving framework designed to deploy models across multiple GPUs in data center environments. The framework handles request routing and communication coordination for models that exceed single-GPU capacity.
The new recipes are optimized for Nvidia B200 GPUs and claim to deliver up to 2x performance improvement out of the box. Nvidia says the recipes reduce deployment complexity for developers working with these Korean models.