NVIDIA Nemotron 3 Ultra Release

NVIDIA Nemotron 3 Ultra is an open model designed for complex, long-running agentic workflows, providing high accuracy and cost-efficiency for tasks involving hundreds of tool calls. It is now available on Ollama's cloud, enabling developers to integrate it into agent orchestration, coding agents, and deep research pipelines.

Model Architecture and Efficiency

Nemotron 3 Ultra utilizes a Mixture-of-Experts (MoE) architecture with 550 billion total parameters, but only 55 billion are active per token. This design allows the model to maintain frontier-level reasoning capabilities while reducing the computational overhead per token.

To further optimize performance and memory footprint, the model is optimized for NVFP4, NVIDIA's 4-bit floating point format. This quantization allows the model to be packed into less memory and execute faster than standard formats.

Long-Context and Agentic Capabilities

Nemotron 3 Ultra is specifically tuned for agent orchestration and enterprise workflows that span hundreds of steps. Key capabilities include:

  • 1M Token Context Window: The model can maintain entire codebases, extensive tool histories, and research trails within its context window without losing coherence.
  • Agentic Specialization: The model is optimized for coding agents, deep research, and complex workflows requiring high-precision instruction following and multiple tool calls.

Performance and Cost Benchmarks

According to NVIDIA and Ollama, Nemotron 3 Ultra leads among open models in accuracy across agent productivity, coding, and instruction following benchmarks.

In terms of operational efficiency, the model demonstrates leading throughput and cost-effectiveness. It saves up to 30% in costs compared to other leading open models, placing it on the leading edge of the cost-efficiency frontier.

Deployment and Integration

Nemotron 3 Ultra can be deployed via Ollama's cloud. It supports integration with several agentic tools, including:

  • Claude Code: ollama launch claude --model nemotron-3-ultra:cloud
  • Hermes Agent: ollama launch hermes --model nemotron-3-ultra:cloud
  • OpenClaw: ollama launch openclaw --model nemotron-3-ultra:cloud
  • General Chat: ollama run nemotron-3-ultra:cloud

Sources

Related

  • Dispatch
    NVIDIA Nemotron 3 Ultra Support on vLLM

    vLLM announces Day-0 support for NVIDIA Nemotron 3 Ultra, a 550B parameter model optimized for long-running autonomous agent workflows through a hybrid Transformer-Mamba MoE architecture.

  • Dispatch
    NVIDIA Nemotron 3.5 Lightning Release

    NVIDIA Nemotron 3.5 Lightning is a 30B parameter open model with 3B active parameters per token, designed for local agentic workflows and multi-step tasks with a 1M token context window.

  • Dispatch
    NVIDIA Nemotron 3 Super Support in vLLM

    vLLM now supports NVIDIA Nemotron 3 Super, a 120B parameter hybrid MoE model optimized for multi-agent AI with a 1 million token context window and high inference efficiency.

  • Dispatch
    NVIDIA Nemotron 3 Nano Omni Support in vLLM

    vLLM now supports NVIDIA Nemotron 3 Nano Omni, a highly efficient 30B MoE multimodal model that unifies vision, audio, and language reasoning in a single loop to power agentic AI.

  • Dispatch
    NVIDIA Nemotron 3 Nano Omni release notes / what's new

    NVIDIA has released Nemotron 3 Nano Omni, an omni-modal model capable of long-context reasoning across text, images, video, and audio, delivering best-in-class accuracy on document intelligence and video understanding benchmarks.