Editorial illustration for Nvidia Launches Open Models for Enterprise AI Agents
Nvidia Launches Open AI Agent Models for Enterprise
Nvidia rolled out a new set of open-weight models on Thursday aimed at a specific problem: companies want AI agents that run constantly, on their own infrastructure, without paying a premium to OpenAI or Anthropic for every task. The headline release is Nemotron 3.5 Super VL, a 120-billion-parameter model built for enterprise agent work. Nvidia says it matches the accuracy of its larger Nemotron 3 Ultra model on instruction-following and tool-use despite the smaller parameter count.
The company also shipped two lightweight Bolt models for the other end of the market: local inference on AI PCs running the new Nvidia RTX Spark chip, and on DGX Spark workstations. Those machines are hitting shelves now through partners including Microsoft and Dell, targeting consumers, developers, and knowledge workers who want agents running on-device rather than in the cloud.
Nvidia is positioning the release as a full-stack play, covering everything from enterprise servers down to a programmer's desktop, so customers don't have to pick between a model that's capable and one that's actually practical to deploy. The company is also making pointed comparisons to rivals' latest models on specific agentic benchmarks, a claim worth unpacking.
“Any organization that wants to run agents on Nvidia infrastructure now has an open, optimized model for their tier without having to choose between capability and deployability,” Nvidia’s vice president of Open Models and Software, Kari Briski, wrote in the blog post announcing these models.
Why this matters
Nvidia isn't just selling chips anymore, it's trying to own the model layer that sits on top of them. By releasing Nemotron 3.5 Super VL and the smaller Bolt models as open weights, the company gives developers a reason to standardize on Nvidia infrastructure end to end, from training to inference to the PCs agents run on locally. Briski's framing, "an open, optimized model for their tier," is really a pitch for lock-in dressed up as flexibility.
For teams building enterprise agents, that's worth watching closely: a 120-billion-parameter model that matches a larger one on instruction-following and tool use is a real engineering claim, and it's worth testing against your own tool-calling benchmarks rather than taking Nvidia's word for it. The Bolt models matter more for anyone building on-device agents for PCs, where latency and privacy constraints rule out cloud calls. Open weights lower the barrier to inspection and fine-tuning, which is good.
But the real question is whether "optimized for your tier" quietly means optimized for Nvidia's stack specifically, and how portable these models actually are off it.
Related Reading
Business & Startups Nordic pilot adds Gemini for Education, NotebookLM to boost AI literacy
Business & Startups Kling launches Video O1, all-in-one model with MVL bridge using transformer
Business & Startups DeepSeek Seeks More Capital Weeks After USD 7B Funding Round
Related Topic NVIDIA and Google Cloud let developers scale AI from prototype to production
Market Trends
Related Topic NVIDIA NeMo powers telco reasoning model for autonomous network workflows
Industry Applications
From Archives MiniMax M3.1 Flash at USD 0.15/M; Kimi K3 scores match table, cost chart diverges
From Archives Karnataka ElevenLabs partnership irks Indian voice AI startups over credibility
From Archives NVIDIA RTX Spark Puts AI Development Entirely on One Machine
AI Tools & Apps
From Archives NVIDIA Nemotron 3.5 Super VL customizes agents for domain data, tools
LLMs & Generative AI
Common Questions Answered
What is Nemotron 3.5 Super VL and how does it compare to Nemotron 3 Ultra?
Nemotron 3.5 Super VL is a 120-billion-parameter open-weight model released by Nvidia specifically designed for enterprise AI agent work. Despite having fewer parameters than the larger Nemotron 3 Ultra model, it matches the accuracy of the Ultra version on instruction-following and tool-use tasks, making it a more efficient option for organizations.
Why did Nvidia release open-weight models for enterprise AI agents?
Nvidia released these models to address a specific enterprise problem: companies want AI agents that can run continuously on their own infrastructure without paying premium fees to providers like OpenAI or Anthropic for each task. By offering open-weight models, Nvidia enables organizations to deploy capable agents cost-effectively on their own systems.
How does Nvidia's strategy with Nemotron models and Bolt models create infrastructure lock-in?
By releasing optimized open-weight models like Nemotron 3.5 Super VL and smaller Bolt models, Nvidia encourages developers to standardize on Nvidia infrastructure across the entire AI pipeline, from training to inference to local PC deployment. This strategy creates end-to-end lock-in while presenting itself as offering flexibility and choice to developers.
What infrastructure advantages does running Nemotron 3.5 Super VL on Nvidia hardware provide?
Nemotron 3.5 Super VL is optimized specifically to run on Nvidia infrastructure, allowing organizations to deploy enterprise AI agents efficiently without paying per-task fees to third-party providers. The model is designed to work seamlessly across Nvidia's hardware ecosystem, from data center training to local PC inference.
Further Reading
- Enterprise Software Leaders Build AI Agents With NVIDIA - NVIDIA
- NVIDIA Debuts Three New Open AI Models - Yahoo Finance
- Open From Data Center to Desktop: NVIDIA Expands Nemotron Family With Highest Efficiency Super VL and Bolt Models - NVIDIA Blog
- NVIDIA Nemotron 3.5 Super VL Powers Specialized Agents With Multimodal Reasoning - NVIDIA Technical Blog
