Both sources agree that Nvidia's new Nemotron 3.5 Super VL is an open-weight, 120-billion-parameter model built for specialized agentic workloads, and that it ships with two smaller edge models: Bolt 74B-A9B 24GB for local agents and Bolt E4B 4GB for conversational, device-based agents. Super VL is available now, with Bolt E4B following on Oct. 23. Nvidia says the lineup covers deployments from large enterprise servers down to local RTX hardware, giving organizations an open model for each tier.
On performance, Nvidia says Super VL is the fastest open agentic model in its class, claiming a 5x speed advantage over the next best; ZDNET attributes that to Artificial Analysis benchmarks and notes that the base model does not top proprietary frontier models on its own. Instead, ZDNET emphasizes post-training gains, citing a 44% improvement in medical insurance claims that moved the model above Claude Opus 5. Nvidia's own release similarly stresses post-training flexibility, open weights, and the ability to run on private infrastructure.
Where the sources differ most is in framing. Nvidia highlights control, security, and adoption across industries, naming partners such as Perplexity, Salesforce, and CrowdStrike. ZDNET places the release in a broader context: US open-weight models are catching up with Chinese labs, and events like the Hugging Face breach are making open models more attractive to enterprises. Both agree the open-weight direction is significant, but ZDNET treats it as a competitive shift rather than just a product launch.