Nvidia Releases 24GB Bolt Model

Nvidia announced the Nemotron 3.5 Bolt 74B-A9B 24GB model on Thursday, squeezing a 74-billion-parameter agent into 24GB via two-bit quantization. The launch targets local execution on RTX Spark and DGX Spark systems for coding and research workloads, with a smaller 4GB version arriving later this month.