Nemotron 3.5 Lightning
NVIDIA's 30B-A3B Mamba-2/MoE hybrid architecture with NVFP4 quantization and millisecond latency.
Updated 8/30/20261 code examples
About the Model
为什么要选择 Nemotron 3.5 Lightning？
Nemotron 3.5 Lightning 是 NVIDIA 于 2026 年 8 月最新发布的轻量级旗舰 MoE 模型。总参数 30B，每次仅激活 3B 参数，融合了 Mamba-2 状态空间模型与 MoE 混合架构，专为高并发 Coding Agent、工具编排与实时自动补全量身定制。
核心优势与技术突破
- 30B-A3B 极速架构：单 Token 推理延迟大幅低于传统 Dense 架构，首字生成时间达到毫秒级，大幅缩短 Agent 决策循环等待时间。
- 原生 NVFP4 深度优化：在 NVIDIA Blackwell / Hopper 架构 GPU 上深度优化，显存开销更小、吞吐速度翻倍。
- 100 万（1M）Token 上下文：支持长文本代码审阅与复杂系统日志分析，长程上下文忠实度极高。
How to Access for Free (via NVIDIA NIM)
Why Choose Nemotron 3.5 Lightning on NVIDIA NIM?
Released in August 2026, Nemotron 3.5 Lightning combines Mamba-2 state-space efficiency with MoE sparsity (30B total, 3B active) for the lowest latency in autonomous agent execution.
Try it in your browser
Pick a model, paste your own free key, and run. Your key is sent once to call the provider and never stored on our servers.
Pick a model, paste your own free key, and run. Your key is sent once to call the provider and never stored on our servers.
Code Examples
curl
curl https://integrate.api.nvidia.com/v1/chat/completions \
-H "Authorization: Bearer $NVIDIA_API_KEY" \
-H "Content-Type: application/json" \
-d '{
"model": "nvidia/nemotron-3.5-lightning-30b-a3b",
"messages": [{"role": "user", "content": "Explain Mamba-2 state space models."}]
}'
Community Feedback
Comments are tied to your GitHub account — sign in to join the discussion.