Nvidia released Nemotron 3.5 Super VL, an open, 120-billion-parameter AI model for enterprise agents, on October 15, and announced two Bolt models for local systems.

Super VL is available on Hugging Face, OpenRouter, and ModelScope, as well as through Nvidia NIM, the company’s microservice for production deployment.

Bolt 74B-A9B 24GB is also available, while Bolt E4B 4GB is due on October 23.

Nvidia said companies can post-train Super VL on their own data and deploy it on their own infrastructure.

It lists Atlassian, CrowdStrike, Salesforce, Perplexity, and Dassault Systèmes as users that are integrating the model for search, security, government workflows, and engineering tasks.

Nvidia also has a public case study describing Justt’s use of Nemotron Parse for chargeback document extraction, not Super VL.

An enterprise model comparison names Qwen 2.5, Llama 3.2, and DeepSeek R1 as alternative AI models for enterprises.

🔗 Source: Nvidia

Recent Nvidia developments