英伟达发布 Nemotron 3.5 Lightning 与 NeMo
NVIDIA AI Releases Nemotron 3.5 Lightning: A 30B Open MoE with 3B Active Parameters, and NeMo Switchyard Model Router
做 Agent 的同学注意了，英伟达这套组合拳直击执行层成本痛点，Lightning 模型加 Switchyard 路由实测能砍掉七成成本，赶紧拿你的工作流压测一下。
NVIDIA introduced open technologies for building always-on AI agents from systems of specialized models. Two artifacts shipped together. Nemotron 3.5 Lightning is a lightweight, customizable open model built for high-volume agentic tasks, and NeMo Switchyard is an open source routing library that directs each step of an agent workflow to the most capable and efficient model available. The problem both address is structural: long-running agents spend most of their time on tool calls, result validation, and subagent delegation, and sending every one of those steps to a frontier reasoning model adds cost and latency. Lightning is a 30B mixture-of-experts model with 3B active parameters, built on a hybrid Mamba-2 + MoE + Attention architecture with a 1M-token context window. NVIDIA reports up to 4x faster output speed than similar-sized models, and 30% faster completion of 10,000 PinchBench tasks than Qwen3.6 35B at comparable accuracy. Many industry players like CrowdStrike, Harvey, CodeRabbit, Fastino Labs, and Lila Sciences are already customizing it for cybersecurity, legal, coding, finance, and healthcare workloads.
NVIDIA 推出了用于从专用模型系统构建常驻 AI 代理的开放技术。两个工件一同发布。Nemotron 3.5 Lightning 是一个轻量级、可定制的开放模型，专为高容量代理任务而构建，而 NeMo Switchyard 是一个开源路由库，可将代理工作流的每一步引导至最强大且最高效的可用模型。两者解决的问题是结构性的：长时间运行的代理将大部分时间花在工具调用、结果验证和子代理委派上，而将每一步都发送给前沿推理模型会增加成本和延迟。Lightning 是一个 30B 的专家混合模型，具有 3B 个激活参数，基于混合 Mamba-2 + MoE + Attention 架构，上下文窗口为 1M 个 token。NVIDIA 报告称，其输出速度比类似规模的模型快 4 倍，在 PinchBench 上完成 10,000 个任务的速度比 Qwen3.6 35B 快 30%，且准确率相当。CrowdStrike、Harvey、CodeRabbit、Fastino Labs 和 Lila Sciences 等许多行业参与者已经在针对网络安全、法律、编码、金融和医疗保健工作负载对其进行定制。
Is it deployable?
它可以部署吗？
Yes. Nemotron 3.5 Lightning is generally available under the permissive OpenMDW-1.1 license, with open weights, training data, and recipes. NVIDIA states the model is ready for commercial use.
可以。Nemotron 3.5 Lightning 在宽松的 OpenMDW-1.1 许可下普遍可用，具有开放的权重、训练数据和配方。NVIDIA 表示该模型已准备好用于商业用途。
- Which companies: Anyone with a single modern GPU. NVIDIA lists single-GPU deployment on 1x DGX Spark (GB10) or 1x H100. That puts solo developers and seed-stage startups on the same footing as enterprises. Mid-market teams can serve it from Baseten, Together AI, or Nebius; regulated enterprises can keep it fully on-premises.
- Industries: Cybersecurity, legal services, software engineering, financial services, healthcare, and life sciences all appear in NVIDIA’s named customer set.
- Applications: Tool calling, result validation, subagent delegation, code review routing, log triage, contract parsing, and long-context retrieval across a 1M-token window.
- 哪些公司：任何拥有单个现代 GPU 的公司。NVIDIA 列出了在 1x DGX Spark (GB10) 或 1x H100 上的单 GPU 部署。这使独立开发者和种子期初创公司与大型企业处于同等地位。中型市场团队可以从 Baseten、Together AI 或 Nebius 提供服务；受监管的企业可以将其完全保留在本地。
- 行业：网络安全、法律服务、软件工程、金融服务、医疗保健和生命科学都出现在 NVIDIA 指定的客户群中。
- 应用：工具调用、结果验证、子代理委派、代码审查路由、日志分类、合同解析以及跨 1M token 窗口的长上下文检索。
The execution layer, not the planning layer
执行层，而非规划层
Long-running agents spend most of their time on high-volume execution. Tool calls, result validation, and subagent delegation dominate the token budget. Routing every one of those steps to a frontier reasoning model adds cost and latency.
长时间运行的代理将大部分时间花在高容量执行上。工具调用、结果验证和子代理委派主导了 token 预算。将每一步都路由到前沿推理模型会增加成本和延迟。
Nemotron 3.5 Lightning targets that execution layer. It is a 30B mixture-of-experts model with 3B active parameters, built on a hybrid Mamba-2 + MoE + Attention architecture. Context length reaches 1M tokens. Pre-training covered more than 20 trillion tokens using an NVFP4 recipe.
Nemotron 3.5 Lightning 针对的是执行层。它是一个 30B 的混合专家模型，拥有 3B 激活参数，基于混合 Mamba-2 + MoE + Attention 架构构建。上下文长度达到 1M tokens。预训练使用了 NVFP4 配方，覆盖超过 20 万亿 tokens。
The model is the smallest member of the Nemotron 3 family. Frontier models such as Nemotron 3 Ultra handle orchestration and planning, while Lightning handles the routine calls beneath them.
该模型是 Nemotron 3 家族中规模最小的成员。像 Nemotron 3 Ultra 这样的前沿模型负责编排和规划，而 Lightning 则处理它们之下的日常调用。
Where the speed comes from
速度的来源
Two mechanisms:
两种机制：
- First, Speculative Decoding: Multi-token prediction was baked in during a dedicated pre-training stage, then improved with an MTP-boosting phase. NVIDIA also ships two external draft models: DSpark, a semi-autoregressive drafter recommended for DGX Spark and low-concurrency data center workloads, and DFlash, which uses a lightweight block-diffusion model.
- Second, Quantization: An NVFP4 checkpoint ships alongside BF16. The same checkpoint serves Blackwell and Hopper natively, and extends to Ampere through W4A16 kernels.
- 第一，推测解码：在专门的预训练阶段就内置了多 token 预测，然后通过 MTP 增强阶段进行改进。NVIDIA 还提供了两个外部草稿模型：DSpark，一个半自回归草稿模型，推荐用于 DGX Spark 和低并发数据中心工作负载；以及 DFlash，它使用轻量级块扩散模型。
- 第二，量化：NVFP4 检查点与 BF16 一起提供。同一检查点原生支持 Blackwell 和 Hopper，并通过 W4A16 内核扩展到 Ampere。
NVIDIA reports up to 4x output speed versus similar-sized models. On PinchBench, it reports 86% accuracy while completing 10,000 tasks 30% faster than Qwen3.6 35B at comparable accuracy.
NVIDIA 报告称，与类似规模的模型相比，输出速度最高提升 4 倍。在 PinchBench 上，它报告了 86% 的准确率，同时完成 10,000 个任务的速度比 Qwen3.6 35B 快 30%，且准确率相当。
Published model card results (BF16 / NVFP4): MMLU Pro 81.94 / 81.62, GPQA Diamond 75.44 / 75.57, SWE-bench Verified 51.56 / 52.80, Terminal-Bench 2.1 24.58 / 23.46, AA-LCR 52.00 / 49.19. Recommended sampling is temperature 1.0 and top_p 0.95.
已发布的模型卡结果（BF16 / NVFP4）：MMLU Pro 81.94 / 81.62，GPQA Diamond 75.44 / 75.57，SWE-bench Verified 51.56 / 52.80，Terminal-Bench 2.1 24.58 / 23.46，AA-LCR 52.00 / 49.19。推荐采样参数为 temperature 1.0 和 top_p 0.95。
NeMo Switchyard
NeMo Switchyard
NeMo Switchyard is an open source library that routes each step of an agent workflow to the most capable and efficient model available.
NeMo Switchyard 是一个开源库，它将代理工作流的每一步路由到最强大且最高效的可用模型。
It offers tuning-free routers, including an LLM classifier with session affinity, a stage router that reads recent tool activity, and an escalation router that starts cheap and promotes on sustained difficulty. A tunable prefill router learns from the model’s residual stream to predict which candidate will succeed. The reference server accepts OpenAI, Anthropic, and Responses API requests.
它提供了无需调优的路由器，包括具有会话亲和性的 LLM 分类器、读取最近工具活动的阶段路由器，以及从低成本开始并在持续困难时升级的升级路由器。可调的前缀路由器从模型的残差流中学习，以预测哪个候选者会成功。参考服务器接受 OpenAI、Anthropic 和 Responses API 请求。
Two published results: LangChain benchmarked 145 multi-turn agentic tasks. Routing between Lightning and Claude Opus 4.8 with the escalation router cut cost 74% versus a frontier-only baseline, sending 7% of calls to the frontier model, at a roughly 6-point accuracy tradeoff. Cognition implemented staged routing in Devin Desktop. On FrontierCode Main, routing between Opus 5 and Kimi K2.7 reached 50.6% at a $3.11 mean cost, within 2.8 points of Opus 5 accuracy at approximately 28% lower mean cost.
两项已发表的结果：LangChain 对 145 个多轮智能体任务进行了基准测试。在 Lightning 和 Claude Opus 4.8 之间使用升级路由器进行路由，相比仅使用前沿模型的基线，成本降低了 74%，将 7% 的调用发送给前沿模型，准确率大约损失 6 个百分点。Cognition 在 Devin Desktop 中实现了分阶段路由。在 FrontierCode Main 上，Opus 5 和 Kimi K2.7 之间的路由达到了 50.6% 的准确率，平均成本为 3.11 美元，与 Opus 5 的准确率相差 2.8 个百分点，平均成本大约降低 28%。
Interactive explainer
交互式解释器
Key Takeaways
关键要点
- 30B open MoE with 3B active parameters, 1M context, OpenMDW-1.1 license, commercial use permitted.
- Up to 4x output speed; PinchBench 10,000 tasks completed 30% faster than Qwen3.6 35B.
- Speed comes from multi-token prediction plus DSpark and DFlash drafters, and an NVFP4 checkpoint.
- Runs on 1x DGX Spark or 1x H100, and locally via Ollama, LM Studio, llama.cpp, and Unsloth.
- NeMo Switchyard cut cost 74% in LangChain’s 145-task benchmark at a ~6-point accuracy tradeoff.
- 30B 开放 MoE，3B 激活参数，1M 上下文，OpenMDW-1.1 许可证，允许商业使用。
- 输出速度最高提升 4 倍；PinchBench 10,000 个任务完成速度比 Qwen3.6 35B 快 30%。
- 速度来自多 token 预测以及 DSpark 和 DFlash 草稿模型，以及 NVFP4 检查点。
- 可在 1 个 DGX Spark 或 1 个 H100 上运行，也可通过 Ollama、LM Studio、llama.cpp 和 Unsloth 在本地运行。
- NeMo Switchyard 在 LangChain 的 145 任务基准测试中降低了 74% 的成本，准确率大约损失 6 个百分点。
Try it on build.nvidia.com or OpenRouter, and download weights from Hugging Face or ModelScope. Also, feel free to follow us on Twitter and don’t forget to join our 150k+ML SubReddit and Subscribe to our Newsletter. Wait! are you on telegram? now you can join us on telegram as well.
在 build.nvidia.com 或 OpenRouter 上试用，并从 Hugging Face 或 ModelScope 下载权重。此外，欢迎在 Twitter 上关注我们，别忘了加入我们的 150k+ ML SubReddit 并订阅我们的新闻通讯。等等！你在 Telegram 上吗？现在你也可以在 Telegram 上加入我们。
Need to partner with us for promoting your GitHub Repo OR Hugging Face Page OR Product Release OR Webinar etc.? Connect with us
需要与我们合作推广你的 GitHub 仓库、Hugging Face 页面、产品发布或网络研讨会等？请联系我们。
The post NVIDIA AI Releases Nemotron 3.5 Lightning: A 30B Open MoE with 3B Active Parameters, and NeMo Switchyard Model Router appeared first on MarkTechPost.
这篇帖子《NVIDIA AI 发布 Nemotron 3.5 Lightning：一个具有 3B 激活参数的 30B 开放 MoE，以及 NeMo Switchyard 模型路由器》首次出现在 MarkTechPost 上。
更进一步：量化金融体系
看懂新闻只是起点——沿量化金融路径，把它变成能交付的工程能力