ZDNET’s key takeaways
- Three new Nvidia open models strive for efficiency and speed.
- Open models are having a renaissance on the heels of the Hugging Face Hack.
- US labs start in the open model race.
China-based AI laboratories have had the upper hand on competitive open weight AI models in recent years, but a recent surge of Western-made open models is challenging that.
Also: Open Weights vs. Closed: An AI Civil War Is
Nvidia’s new model, the open-weight Nemotron 3.5 Super VL, aims to be the latest data point in that stream. The 120-billion-parameter model, available today, addresses the industry’s increased focus on agentic work by “delivering the fastest task completion and highest accuracy for specialized, always-on agents,” according to Nvidia.
“As the fastest open agent model in its class—5x faster than the next best—it helps enterprise agents complete high-volume tasks faster across workflows that span days or weeks,” Nvidia said in the release. This is based on artificial analysis benchmark testing of the model against its similar major competitors; InclusionAI’s Ling 3.0 Flash and DeepSeek’s V4.1 Flash came in second and third, respectively, but with considerable distance between them and Super VL.
Nvidia also said that Super VL achieves top-tier accuracy for specialized agents with post-training. The company cites how far post-training a basic version of the model could take its capabilities – in one case, a gain of 44% in medical insurance claims, which the model far exceeds Claude Opus 5. While this means that the basic model does not have top proprietary Frontier models alone, it gives companies the chance to tailor the model to their needs and see highly specific gains.
As Nvidia pointed out in the release, the advantage open models have over proprietary models is that businesses can post-train them on their own data and run them on their own infrastructure, which provides more personalization and security.
Super VL is also flanked by two new edge models: Nemotron 3.5 Bolt 74B-A9B 24GB and Nemotron 3.5 Bolt E4B 4GB (no, Nvidia didn’t waste time to mention here). Edge AI refers to models that can operate locally and off-grid. Nvidia said the former Bolt is built for local agents, while the latter Bolt is for “fast, conversational agents.”
US Open-Weight AI begins
Although it is not clear whether Nvidia’s new model will raise the bar for open weight or proprietary models in practice, the findings of Artificial Analysis show efficiency gains for now.
Also: I tested Gemini Spark – another Google AI failure, or its next dark horse?
In the bigger picture, however, the release is part of a broader shift in the AI world. Open models from Chinese companies like DeepSeek and Moonshot.ai have historically embodied the promise (and risk) of open, cheaper, more customizable models compared to closed, expensive ones. But the French lab Mistral has just launched its Le Chonk model, which has achieved certain state-of-the-art benchmarks. Reflection AI, which Nvidia has invested in and is in talks to acquire, released its Beam model last week, also to much fanfare.
The geography gap can close in the open weight wars.
The industry is also embracing open weight models more than before, after the OpenAI Hugging Face breach, which dispelled the previous notion that only open models pose security risks. If US-made open models prove themselves against Chinese, this could help to open open models in areas where they do not have yet, giving cybersecurity guidance against the use of foreign tech.
See full bio