Nvidia has launched a new set of open AI models designed to run always-on agents both in enterprise data centers and directly on AI-enhanced PCs. Their approach targets companies wanting customizable on-premise AI as well as users seeking privacy-focused local assistants.
- Nemotron 3.5 Super VL delivers enterprise-scale agentic AI with 120B parameters
- Bolt models run AI assistants locally on new Nvidia-powered PCs for privacy
- Open-weight models enable organizations to customize and control AI deployment
What happened
Nvidia introduced a new suite of AI models tailored for both enterprise and consumer use cases. The Nemotron 3.5 Super VL is a large, open-weight model with 120 billion parameters optimized for enterprise agents performing tasks like document intelligence, financial accounting, and security response. Alongside this, Nvidia released two lightweight Bolt models designed to run locally on AI PCs powered by the Nvidia RTX Spark chip and DGX Spark workstations. These smaller models support local execution of AI assistants, focusing on speed, reasoning, and conversational capabilities.
These launches coincide with Nvidia's strategy to support enterprises that prefer to run AI models internally on their own infrastructure rather than relying on cloud-hosted, closed-source models. The company highlights user organizations such as Skydance, EY, and Salesforce already adopting Super VL. Additionally, Nvidia’s Bolt models aim to address privacy concerns by enabling consumers and enterprise knowledge workers to run private AI agents on-device without transmitting sensitive data to the cloud.
Why it matters
Nvidia’s new open-weight models emphasize the growing industry demand for AI models that companies can customize and control on-premises. By providing competitive accuracy at high speed and compatibility with Nvidia’s hardware ecosystem, these models aim to reduce reliance on proprietary cloud AI services and foster greater data privacy and security for enterprise users.
For consumers and knowledge workers, the Bolt models represent a step toward AI assistants that operate fully locally, alleviating privacy fears associated with cloud-based AI interaction. Nvidia’s approach could accelerate the adoption of AI in everyday workflows while supporting a new wave of PC hardware designed specifically for AI workloads, in partnership with major manufacturers like Microsoft and Dell.
What to watch next
Attention will focus on how widely enterprises adopt the Super VL model as a replacement or complement to closed-source AI offerings from frontrunners like OpenAI and Anthropic. The ability for organizations to post-train the open model on proprietary data and deploy it securely on Nvidia infrastructure could become a critical differentiator in regulated industries.
For consumer AI, the reception of Bolt-powered PCs by privacy-conscious users will be key. As Bolt E4B model availability expands later this month, market feedback on real-world performance and usability will shape Nvidia’s momentum in creating a local AI assistant ecosystem. Additionally, partnerships with OEMs and software developers to deliver integrated AI experiences will influence how fast this segment grows.