The new open‑source Nemotron 3.5 Super VL and its lightweight Bolt variant are designed to power enterprise AI agents both in the cloud and on next‑generation AI‑enabled personal computers.

Nvidia has unveiled two new open‑source language models—Nemotron 3.5 Super VL and its streamlined Bolt variant—targeted at enterprise AI agents running in data‑center clouds and on next‑generation AI‑enabled personal computers.

Nemotron 3.5 Super VL: A heavyweight for data‑center agents

The Super VL model expands Nvidia’s Nemotron line with a larger context window and enhanced multimodal capabilities, allowing agents to process longer documents, richer visual inputs, and more complex reasoning tasks without sacrificing latency.

Built on the same transformer architecture as earlier Nemotron releases, Super VL incorporates Nvidia’s latest tensor core optimizations, delivering up to 2.5× higher throughput on A100‑based clusters compared with the prior 3.0 version.

Bolt: A lightweight model for edge PCs

Bolt is engineered for AI‑enabled desktops and laptops, offering a compact footprint that can run on consumer‑grade GPUs such as the RTX 4090 while still supporting real‑time agent interactions.

Despite its reduced parameter count, Bolt retains core language understanding and can execute task‑oriented agents for productivity tools, code assistance, and on‑device privacy‑preserving inference.

  • Optimized for Nvidia’s CUDA and TensorRT runtimes
  • Supports both FP16 and INT8 inference paths
  • Available under an Apache 2.0 license for unrestricted commercial use

Implications for enterprise AI deployments

Enterprises can now mix and match models based on workload demands—deploying Super VL for heavy‑duty analytics in the cloud while leveraging Bolt for low‑latency, on‑premise assistance on employee workstations.

The open‑source licensing also lowers barriers for third‑party developers to fine‑tune the models for domain‑specific agents, potentially accelerating adoption across sectors such as finance, healthcare, and manufacturing.

For more context on Nvidia’s strategy, see Nvidia on Momentum and explore related stories in the Technology section.

Read the full announcement in TechCrunch coverage of Nvidia’s new open models.