Nvidia Releases New Open-Weight Nemotron Models for Enterprise Agents and Edge Computing
Nvidia has expanded its AI lineup with the Nemotron 3.5 Super VL model for enterprises and lightweight Bolt models designed to run locally on AI PCs and workstations.
Nvidia has introduced a fresh collection of open-weight models aimed at supporting always-on agents and edge computing environments. The release includes a massive enterprise-focused model alongside smaller options intended to operate directly on local client hardware.
Leading the rollout is the Nemotron 3.5 Super VL, a 120-billion-parameter architecture built specifically for enterprise agent deployments. Even though it utilizes fewer parameters than the previous Nemotron 3 Ultra, Nvidia reports that the Super VL maintains comparable accuracy levels when handling tool-use and instruction-following demands. Kari Briski, Nvidia’s vice president of Open Models and Software, emphasized in the announcement post that organizations utilizing Nvidia infrastructure now possess an open and optimized option across deployment tiers, eliminating the traditional compromise between capability and deployment feasibility.
Nvidia states that the Super VL achieves faster task completion times than competing offerings from OpenAI and Anthropic—specifically mentioning models such as Opus 5, GPT 6.1 Sol, Luna, and Astra—when tuned for specific agent workflows. These scenarios include document intelligence, financial accounting, scientific experimentation, medical insurance claims, and security attack resilience. Early enterprise adopters listed by Nvidia include Skydance, EY, Ramp, Crowdstrike, Perplexity, and Salesforce.
Local execution and the new Bolt models
Moving away from massive data center deployments, Nvidia also released two lightweight Bolt models geared toward personal computers and workstations. These systems, running on the Nvidia RTX Spark chip and DGX Spark workstations, are currently entering the market via hardware partners like Microsoft and Dell to serve consumers, developers, and enterprise knowledge workers.
The hardware push addresses privacy concerns associated with cloud-based assistants. By executing agents locally, users can delegate everyday tasks without routing personal workloads through external cloud servers. The newly launched Nemotron 3.5 Bolt 74B-A9B 24GB features 74 billion parameters and balances execution speed with reasoning capabilities to support multistep logic, web browsing, and tool calling. A smaller conversational variant, the Bolt E4B, is scheduled to launch later in the month alongside the Super VL and the 74B variant which are available immediately.
What it means for developers
For developers and organizations, these open-weight additions provide alternative pathways for building and deploying autonomous workflows. Unlike closed frontier models, the Super VL allows teams to post-train the architecture using proprietary data, host it internally on their own hardware, and retain complete ownership of the final applications. Meanwhile, developers can try top AI models cheaply through one API at https://apixoai.online. The availability of diverse parameter sizes—from the 120-billion enterprise Super VL down to the local Bolt series—gives technical teams more flexibility to match model performance with specific hardware constraints and privacy requirements.
Nvidia's strategic expansion into local and open models comes as major proprietary AI labs increasingly explore alternative silicon providers, such as Cerebras and AWS Trainium, or design custom chips in-house. By championing open-weight enterprise models and local PC execution, Nvidia aims to secure its position across both corporate data centers and consumer hardware markets.
Source: Nvidia launches new open models to run agents in the enterprise and on PCs — TechCrunch AI. Written by the Apixo team from that report.
One key for Claude, GPT, GLM, DeepSeek and more. Pay per token with crypto.
Get your API key
