NVIDIA expands Cosmos synthetic-data, training and simulation tools

NVIDIA says its Cosmos expansion adds training tools, simulation checkpoints and five Data Factory blueprints, including synthetic driving scenarios and industrial defects. For developers, NVIDIA’s described workflow connects dataset preparation, model adaptation and simulated policy evaluation before real-world testing—not a replacement for that testing. Source

NVIDIA says new Cosmos 3 Nano simulation variants support repeatable, multiview, multisensor scenarios for evaluating autonomous-vehicle and bimanual-robot policies. Whether those evaluations reliably predict real-world behavior remains unverified. Source

NVIDIA reports up to twice-as-fast inference from updated Cosmos 3 NIM microservices, but limits its no-quality-loss claim to validated configurations. NVIDIA also describes Porsche’s potential reductions in test drives and development time as expectations, not achieved results. These vendor claims do not independently establish performance across other hardware, workloads or deployment environments. Source