Nemotron 3.5 models unveiled: Open-source MoE hybrids for agentic AI

Researchers have introduced Nemotron 3.5 Super and Lightning, new open-source Mixture-of-Experts hybrid Mamba-Transformer models. Nemotron 3.5 Super is a 120B-parameter multimodal model designed for reasoning and agentic tasks, while Lightning is a smaller 30B-parameter text model focused on efficiency. Both models incorporate advanced training techniques, including large-scale distillation, long-context training, and speculative decoding, with Super also featuring specialized agentic training for software engineering capabilities. The Super model demonstrates strong performance on benchmarks like SWE Bench Pro, outperforming its predecessor Nemotron 3 Ultra despite having fewer active parameters, and achieves high inference speeds on GB200 hardware. AI

IMPACT These open-source models could accelerate research and development in agentic AI and efficient model deployment.

RANK_REASON The item is a research paper detailing new open-source AI models. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.AI →

• GB200
• Mamba-Transformer
• mixture of experts
• Nemotron 3.5 Lightning
• Nemotron 3.5 Super
• Nemotron 3 Ultra
• reinforcement learning from human feedback
• RLVR
• SWE Bench Pro

• model release
• product
• infra

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →