Nvidia Launches New Nemotron AI Models
The technology giant releases two high-capacity AI models designed for advanced reasoning and enterprise operations.
Updated on Oct. 19, 2026 in Artificial Intelligence
Live Poll
Would you prefer to run AI models on your own local hardware instead of the cloud?
Nvidia has officially released the Nemotron 3.5 Bolt 74B-A9B and the Nemotron 3.5 Super VL artificial intelligence models. These new releases provide developers and enterprise users with enhanced capabilities for handling complex, multistep reasoning tasks.
Why it matters
The Bolt model is built specifically for agents tasked with managing software tools, web browsing, and reasoning, while the Super VL model targets enterprise workflows involving visual documents. By offering open-weight access, these models allow developers greater control over deployment and adaptation.
The Bolt 74B-A9B model contains 74 billion parameters and requires 24GB of memory to operate. The Super VL model features 120 billion parameters and is optimized for deployment on hardware like Nvidia RTX Spark and DGX Spark workstations.
The players
Nvidia
Nvidia is a global leader in hardware and software development for artificial intelligence, accelerated computing, and data center operations.
Ramp
Ramp is a financial technology company that provides corporate card and spend management software solutions.
CrowdStrike
CrowdStrike is a prominent cybersecurity company specializing in cloud-delivered endpoint protection and threat intelligence.
The details
Running the Bolt model locally allows users to execute AI workloads directly on their hardware without transmitting sensitive data to the cloud. Developers can now adapt the open-weight models to suit specific operational requirements, moving beyond traditional cloud-only AI deployments.
Timeline
October 15, 2026: Nvidia released the Bolt 74B-A9B and Super VL models.
October 2026: The smaller Bolt E4B version is scheduled for release.
The Tech Race
This release reflects the broader industry transition toward edge AI, where heavy workloads move from centralized data centers to local user hardware. By enabling local execution of large models, Nvidia positions itself to capture market share from cloud-dependent providers.
Users can now leverage powerful AI agents for software and web tasks while keeping their data local and secure. This reduces the need for constant cloud connectivity and potentially improves workflow speeds for developers utilizing compatible workstations.
The takeaway
The move toward open-weight models empowers developers to create specialized AI tools that run locally without compromising data privacy. Readers should monitor their hardware specifications to ensure they meet the 24GB memory requirement for local model deployment.
What happens next
Nvidia is scheduled to release a smaller version of its AI model, the Nemotron 3.5 Bolt E4B, later in October 2026.
Further reading
Learn more about the latest developments in Artificial Intelligence.
Live Poll
Would you prefer to run AI models on your own local hardware instead of the cloud?
