NVIDIA today announced Nemotron 3.5 Bolt 74B, a large language model designed to run local AI agents on upcoming RTX Spark laptops. It uses a mixture-of-experts architecture with 74 billion parameters. The model activates 9 billion parameters and is optimized to balance speed and intelligence within a 24GB memory footprint. The release is part of a broader software ecosystem update ahead of the RTX Spark launch on October 16.
In a separate announcement at SEMICON West, NVIDIA said semiconductor industry leaders are adopting its CUDA-X libraries, including the new cuPotential library. cuPotential is designed to accelerate simulations of atomic and molecular interactions, which are important for modeling semiconductor materials and manufacturing processes. The initiative aims to advance the development of AI-defined fabrication plants.