NVIDIA Nemotron 3 Ultra now available on Amazon SageMaker JumpStart

2026-06-04 · Source: Artificial Intelligence · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Cloud Computing & IT Infrastructure, Robotics & Autonomous Systems · Depth: Intermediate, short

Summary

NVIDIA Nemotron 3 Ultra, an open large language model, is now available on Amazon SageMaker JumpStart, offering one-click deployment. This model features a hybrid Transformer-Mamba Mixture-of-Experts (MoE) architecture with 550 billion total and 55 billion active parameters, optimized for the NVFP4 format. It is purpose-built for frontier reasoning and orchestration in long-running autonomous agents, delivering 5x faster inference and up to 30% lower cost for agentic workloads. Nemotron 3 Ultra supports context lengths up to 1 million tokens, enabling sustained multi-step reasoning for tasks like coordinating sub-agents, generating and debugging code, synthesizing information for deep research, and automating complex enterprise workflows. Its MoE design activates only 55 billion parameters per forward pass, maintaining high throughput and managing costs effectively for multi-turn agentic tasks.

Key takeaway

For AI Engineers building complex, multi-step agentic applications, you should consider deploying NVIDIA Nemotron 3 Ultra on Amazon SageMaker JumpStart. Its MoE architecture and NVFP4 optimization deliver 5x faster inference and up to 30% lower costs for sustained reasoning tasks. This enables more efficient development of agent orchestrators, coding agents, and deep research systems. Remember to delete your SageMaker endpoint after use to avoid continuous GPU instance charges.

Key insights

NVIDIA Nemotron 3 Ultra, an MoE model, optimizes cost and speed for complex, long-running AI agent workflows.

Principles

• MoE architectures reduce compute cost for frontier intelligence.
• Agentic AI requires models optimized for multi-turn reasoning.
• NVFP4 format enhances inference speed and cost efficiency.

Method

Deploy Nemotron 3 Ultra via SageMaker JumpStart using one-click deployment or the Python SDK, selecting appropriate GPU instances (e.g., ml.p5en.48xlarge).

In practice

• Use for agent orchestrators managing multi-step tool-calling chains.
• Implement for coding agents to generate, test, and debug code.
• Apply to deep research for synthesizing information over extended contexts.

Topics

• NVIDIA Nemotron 3 Ultra
• Amazon SageMaker JumpStart
• Mixture-of-Experts
• AI Agents
• Large Language Models
• NVFP4

Best for: AI Engineer, Machine Learning Engineer, MLOps Engineer

Related on AIssential

• NVIDIA Nemotron 3 Nano 30B MoE model is now available in Amazon SageMaker JumpStart — NVIDIA's Nemotron 3 Nano 30B, an open-source MoE model, offers high efficiency and accuracy for agentic tasks on AWS.
• NVIDIA Nemotron 3 Ultra Powers Faster, More Efficient Reasoning for Long-Running Agents — Nemotron 3 Ultra optimizes long-running AI agents with a specialized Mixture-of-Experts model for faster, more cost-effective reasoning.
• Nvidia’s Open Salvo, OpenAI’s Amazon Deal, Grok Cuts Video Prices, Recursive Language Models — Nvidia's Nemotron 3 Super offers a fast, open-source LLM for agents, leveraging hybrid architecture and co-designed hardware-software optimization.
• Run NVIDIA Nemotron 3 Nano as a fully managed serverless model on Amazon Bedrock — Nemotron 3 Nano offers an open, efficient, and accurate SLM for specialized agentic AI systems on Amazon Bedrock.
• NVIDIA Releases Nemotron 3 Super: A 120B Parameter Open-Source Hybrid Mamba-Attention MoE Model Delivering 5x Higher Throughput for Agentic AI — Nemotron 3 Super is an open-source 120B parameter hybrid MoE model for efficient multi-agent reasoning.
• Nvidia Nemotron 3 Nano Omni - First Test and Impression — Nvidia's Nemotron 3 Nano Omni is a fast, multimodal MoE model for local inference and diverse data processing.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Artificial Intelligence.