Fine-tune NVIDIA Nemotron 3 models with Amazon SageMaker AI serverless model customization

This article announces serverless model customization for NVIDIA Nemotron 3 models on Amazon SageMaker AI, enabling enterprises to fine-tune open-weight LLMs without managing infrastructure.

• Nemotron 3 uses hybrid Mamba-Transformer Mixture-of-Experts architecture with up to 1M-token context lengths
• Supports three fine-tuning techniques: Supervised Fine-Tuning (SFT), Reinforcement Learning with Verifiable Rewards (RLVR), and Reinforcement Learning from AI Feedback (RLAIF)
• Nemotron 3 Nano (30B total, 3B active) optimized for cost-efficient agentic tasks; Super (120B total, 12B active) for complex reasoning
• SageMaker AI handles compute provisioning, training orchestration, and metric tracking automatically
• Deploy fine-tuned models directly to SageMaker Inference endpoints or download weights for self-managed deployment
• Includes built-in evaluation methods: LLM-as-a-Judge, Custom Scorer, and standardized benchmarks

Serverless customization enables enterprises to create proprietary, domain-specific AI models while reducing infrastructure complexity and operational overhead.