NVIDIA Nemotron 3.5 Lightning now available in Amazon SageMaker JumpStart

This article announces the availability of NVIDIA Nemotron 3.5 Lightning in Amazon SageMaker JumpStart, a specialized model optimized for high-volume agentic workloads.

• Delivers up to 4x higher throughput and 30% faster task completion for agent workflows
• Uses hybrid Mixture-of-Experts architecture with 30B total parameters but only 3B active per forward pass
• Supports 1M-token context window for long-running sessions without re-grounding
• Includes DFlash speculative decoding to reduce per-token latency
• Deploy directly from SageMaker JumpStart, Hugging Face, or SageMaker Python SDK without manual infrastructure configuration
• Open model available in BF16 and NVFP4 variants for customization and domain-specific fine-tuning
• Ideal for personal assistants, financial services, cybersecurity, telecom, and retail use cases

Nemotron 3.5 Lightning enables cost-effective agent execution by handling specialized, high-frequency tasks without requiring frontier-model infrastructure.