Nvidia Unveils Nemotron 3.5 Open AI Models for On-Premise Corporate Assistants

Introducing Nvidia’s Latest AI Innovations

According to НВ — Техно: On October 20th at 10:33, Nvidia launched three new open AI models under the Nemotron 3.5 series, designed specifically for corporate agents and on-premise assistants running locally on computers. These models operate without needing to send data to the cloud, enhancing data privacy and control. The flagship model, Nemotron 3.5 Super VL, boasts 120 billion parameters and is tailored for enterprise use. After fine-tuning, it delivers faster performance than comparable offerings from Anthropic and OpenAI.

Model Details and Capabilities

The lineup also includes two smaller variants:

• Nemotron 3.5 Bolt 74B-A9B with 74 billion parameters;
• Nemotron 3.5 Bolt E4B, focused on building conversational AI assistants.

The Bolt 74B-A9B model excels in multi-step workflows, integrating external tools, and web page browsing. All three models have been optimized to run efficiently on Nvidia’s RTX Spark AI chips and DGX Spark workstations.

Several companies, including Skydance, EY, Ramp, CrowdStrike, Perplexity, and Salesforce, have already started deploying the Nemotron 3.5 Super VL. Kari Briski, Nvidia’s Vice President of Open Models and Software, highlighted the advantage of on-premise deployment:

"Organizations can further fine-tune Super VL using their own data and run the model within their own infrastructure, granting them greater control over solution development and data handling." – Kari Briski

The Bolt E4B model is scheduled for release by the end of October 2026.

Nvidia’s initiative addresses the growing demand from businesses for locally hosted AI models that can operate within their private data centers. These new AI solutions have broad applications across industries, including:

• document management;
• financial accounting;
• scientific research;
• processing medical insurance claims;
• cybersecurity resilience testing.

The introduction of the Nemotron 3.5 series reflects a wider industry shift toward implementing AI models on-premises to meet increasing data privacy and security requirements. This approach empowers organizations to customize AI capabilities to their unique needs while maintaining full control over sensitive information, potentially transforming data workflows across multiple sectors.

In addition to its new AI models, Nvidia is also advancing safety measures for autonomous systems. The recently unveiled Open Agent Safety Platform promises to enhance the isolation of AI agents, ensuring that corporate deployments remain secure and efficient. This initiative further exemplifies Nvidia's commitment to providing robust solutions for businesses navigating the complexities of AI integration.