NVIDIA announced on October 15, 2026 the Nemotron 3.5 Super VL, an open-weight multimodal model designed for long-running agents. It combines reasoning, instruction following, tool use and understanding of images, documents and video in a single architecture.
What happened?
Nemotron 3.5 Super VL is a Mixture-of-Experts with 120 billion total parameters and 12 billion active per token. It supports up to 1 million token context and accepts text, images, documents and video as input, generating text. Weights are available on Hugging Face in BF16 and NVFP4 versions, along with datasets, training recipes and reinforcement learning environments.
NVIDIA claims the model delivers accuracy comparable to its largest previous reasoning model at roughly one-fifth the parameters, and up to 5x higher throughput on agentic tasks in class. Accompanying it are the Bolt models: Nemotron 3.5 Bolt 74B-A9B 24GB (for local agents in 24 GB) and Bolt E4B 4GB (expected October 23, aimed at voice assistants and embedded interfaces).
Super VL runs on NVIDIA hardware from DGX Spark and RTX to data centers. Companies such as Scale AI, Ramp, AI21 and others are already post-training the model for specialized workflows.
Why it matters
High-capacity open-weight models continue lowering the barrier for custom agents that process documents, visual interfaces and tools without relying exclusively on closed APIs. The combination of efficiency (few active parameters) and native multimodality facilitates on-premises or edge deployments where privacy and latency matter.
The Nemotron line expands with options for different scales: Super VL for dense agentic workloads, Bolt for local hardware and prior variants like Lightning and Ultra for other profiles. This reinforces NVIDIA's strategy of offering open models optimized for its hardware stack.
Sources
- NVIDIA Blog — Open From Data Center to Desktop: NVIDIA Expands Nemotron Family
- Hugging Face — NVIDIA-Nemotron-3.5-Super-VL-120B-A12B
- NVIDIA Technical Blog — Nemotron 3.5 Super VL Powers Specialized Agents
Transparency: This content was created, edited, or reviewed with the aid of artificial intelligence. Information was cross-checked with public posts on X and sources available on the internet. Consult the original sources for the full context.
By GeekikiBot
