NVIDIA Releases Open Agentic Models Spanning Data Center to Edge
NVIDIA released Nemotron 3.5 Super VL, a 120-billion-parameter open model designed for enterprise agentic workloads, alongside two edge-optimized variants, Bolt 74B-A9B 24GB and Bolt E4B 4GB. Super VL claims 5x faster throughput than comparable models while delivering accuracy comparable to NVIDIA's largest reasoning models at one-fifth the parameter count. The three models span deployment tiers from data centers to local devices, allowing organizations to run agents on their own infrastructure without trading capability for deployability.
TL;DR
- NVIDIA released Nemotron 3.5 Super VL (120B parameters) for enterprise agentic tasks, claiming 5x faster throughput than competing models in its class
- Two edge models launched: Bolt 74B-A9B 24GB for local agentic workloads on RTX/DGX Spark, and Bolt E4B 4GB for conversational agents on GeForce RTX and Jetson
- Super VL and Bolt 74B-A9B available now, Bolt E4B launching Oct. 23
- Dozens of organizations already post-training Super VL, including Scale AI for Skydance and EY, Mercor for coding agents, and others for cybersecurity and financial intelligence
Why It Matters
Open agentic models reduce vendor lock-in and allow enterprises to customize AI agents on proprietary data without cloud dependencies. Super VL's claimed efficiency gains matter because agentic workloads at scale require both speed and accuracy, and the model's performance claims suggest a meaningful shift in what's possible with open weights at this parameter scale.
Business Impact
Organizations can now deploy specialized agents locally or on-premises without sending sensitive workflows to cloud providers, reducing latency and compliance risk. The range of models from 4GB to 120B parameters means teams can choose the right tool for their infrastructure and performance requirements without overprovisioning.
Key Implications
- Open agentic models are becoming viable alternatives to closed frontier models for enterprise use cases, reducing dependence on API-based services
- Edge deployment of agents is moving from theoretical to practical, enabling local processing of sensitive business workflows like financial analysis and cybersecurity
- Post-training and customization of open models is becoming standard practice across industries, shifting the competitive advantage from base model capability to domain-specific fine-tuning
What to Watch
Monitor adoption rates and performance benchmarks as organizations deploy these models in production. Track whether the claimed 5x throughput advantage holds across real-world agentic workloads and whether edge models like Bolt E4B gain traction in embedded and voice assistant applications. Watch for competitive responses from other open model providers and whether closed model vendors adjust pricing or capabilities in response.
Subscribe to the newsletter
The latest stories and analysis, delivered to your inbox.
Free. No spam. Unsubscribe any time.
