Nvidia Launches High-Speed Nemotron 3.5 Super VL AI Model
The 120-billion parameter model enables high-volume enterprise automation and local data processing.
Updated on Oct. 15, 2026 in Artificial Intelligence
Live Poll
Do you believe the development of U.S.-based open-weight AI models will improve national security and technology?
Nvidia has released the Nemotron 3.5 Super VL, a 120-billion parameter open-weight AI model designed for enterprise agentic workflows. Alongside the primary model, the company introduced two smaller edge-focused models, the Bolt 74B-A9B and Bolt E4B.
Why it matters
These models target the need for high-volume, automated agentic tasks in enterprise environments, allowing organizations to post-train base models on proprietary data for improved specialized performance.
The Nemotron 3.5 Super VL features 120 billion parameters, while the Bolt edge models offer 24GB and 4GB memory capacities respectively. Artificial Analysis benchmark testing demonstrates the Super VL model is 5 times faster than its nearest competitors.
The players
Nvidia
A semiconductor and software company that provides GPU architectures, AI development platforms, and enterprise-grade machine learning models.
Artificial Analysis
A research entity that provides comparative benchmark testing and performance analysis for large-scale AI models.
The details
Enterprises can post-train the base models on their own internal datasets to optimize for specific workflows, such as the 44% accuracy gain observed in medical insurance claim processing. These models are designed to operate on local infrastructure, allowing sensitive enterprise data to remain within a company's secure environment during execution.
Timeline
Nvidia released the Nemotron 3.5 Super VL model on October 15, 2026.
The Tech Race
Nvidia's release of the Nemotron 3.5 series escalates the competition among open-weight AI model providers. The model directly challenges performance benchmarks previously led by competing research labs globally.
Enterprises can deploy these models on their own local infrastructure for tasks like medical insurance claims, where post-training has demonstrated a 44% accuracy improvement. Technical teams will need to manage the specific hardware requirements for the 24GB and 4GB capacity edge models.
The takeaway
Nvidia is pushing to capture the enterprise market by emphasizing local, private infrastructure for high-speed agentic tasks. Watch for future benchmark updates from Artificial Analysis to see if the speed gains persist as other labs update their base models.
Further reading
Explore the broader Artificial Intelligence landscape to understand how enterprise model adoption is changing.
Live Poll
Do you believe the development of U.S.-based open-weight AI models will improve national security and technology?
