NVIDIA drops a massive open bundle: Nemotron models, 10T tokens, and an LLM Router blueprint.

On June 22 NVIDIA released open models, datasets, and tools across the Nemotron family, including training data for the Llama Embed Nemotron 8B embedding model and an updated LLM Router blueprint that auto-directs requests to the best model for a job. (NVIDIA) The broader contribution spans 10 trillion language tokens, 500K robotics trajectories, and 100TB of vehicle sensor data on GitHub and Hugging Face. The Router blueprint is the practical piece. Model routing by task is becoming standard architecture, and NVIDIA just handed you a reference implementation.