Nvidia launches Nemotron 3.5 Super VL open-weight model

By Splitfeed EditorialLast updated Oct 16, 2026, 01:10 UTC

On October 15, 2026, Nvidia released Nemotron 3.5 Super VL, an open-weight model carrying 120 billion parameters. Nvidia described it as the fastest open-weight model at the 120 billion parameter scale. The launch was not a single release but part of a broader update to the Nemotron family, with Bolt variants also announced, bringing the total number of new models released that day to three.

The release matters because it marks a more deliberate push by Nvidia into the open-weight model space, an area that has been dominated by other labs and independent research groups. Open-weight models are publicly available for download and modification, which makes them a distinct competitive category from proprietary API-based systems. For Nvidia, which already controls a dominant share of the hardware used to train and run AI models, entering the open-weight model space directly raises questions about how the company intends to position itself relative to both closed-model providers and the open-source community. Whether the speed claim at the 120 billion parameter scale holds up against independent benchmarks, and what trade-offs exist in quality or capability, remains to be tested publicly.

No reactions from the Pro-AI, Anti-AI or Middle Ground camps had been published at the time of this analysis. Typically, the Pro-AI camp would welcome a powerful open-weight release as democratizing access to frontier-scale models. The Anti-AI camp would likely raise concerns about Nvidia releasing a capable 120 billion parameter model with few restrictions, given ongoing debates about the risks of open-weight systems at large scales. The Middle Ground camp would probably focus on whether Nvidia's efficiency and speed claims are independently verifiable and what safeguards, if any, accompany the release.

The argument that would settle much of this is an independent benchmark evaluation of Nemotron 3.5 Super VL against comparable open-weight models at similar parameter counts. Community testing on platforms that host open-weight models would likely surface within days of the release and provide the first real signal of whether Nvidia's speed and efficiency claims translate outside of controlled conditions.