NVIDIA Releases Nemotron 3.5 Lightning Open Model
Open 30B MoE model with 3B active parameters targets always-on agents.
NVIDIA announced updated weights for its Nemotron 3.5 Lightning model. Director of Research Pavlo Molchanov stated the changes apply distillation from larger Super and Ultra models during continuous pretraining and post-training. Bryan Catanzaro described the result as combining the earlier Nano architecture with Super-level intelligence plus speculative decoding. The open model is positioned for high-volume specialized agent tasks and appears on platforms including Together AI and Prime Intellect.
Introducing NVIDIA Nemotron 3.5 Lightning⚡ An open 30B MoE model with 3B active parameters, built for always-on agents to complete high-volume, specialized tasks faster. It delivers up to 4x the output speed of similar-sized models.
Reactions
72 postsDE 
DeepInfra@DeepInfraDay 0: @nvidia Nemotron 3.5 Lightning is live on DeepInfra. The fastest open model in its class — 30B MoE, 3B active params, 1M context, up to 4x higher throughput for always-on agents. $0.05/M in · $0.20/M out, OpenAI-compatible. More details in technical blog: Try the model…
- likes: 12
- replies: 2
- bookmarks: 2
- reposts: 2
CR Crusoe@CrusoeAINVIDIA Nemotron 3.5 Lightning is now live for inference on Crusoe Intelligence Foundry. Not every agent task needs a frontier model. NVIDIA Nemotron 3.5 Lightning, a 30B MoE, 3B active parameters, delivers the speed and efficiency needed for high-volume, specialized tasks in…
- likes: 30
- replies: 2
- bookmarks: 2
- reposts: 7
NA NVIDIA AI@NVIDIAAILightning pairs strong accuracy with speed. On PinchBench, it reaches 86% accuracy while completing 10,000 tasks 35% faster than Qwen3.6 35B at similar accuracy.
- likes: 303
- replies: 5
- bookmarks: 27
- reposts: 12
NA NVIDIA AI@NVIDIAAILightning is built to specialize. Post-train Nemotron 3.5 Lightning with NVIDIA NeMo for your domain data, tools, workflows and policies. Across cybersecurity, coding, legal and energy tasks, post-training improves accuracy for specialized work.
- likes: 264
- replies: 5
- bookmarks: 66
- reposts: 14
NA NVIDIA AI@NVIDIAAILong-running agents spend most of their time executing: calling tools, validating results and delegating work. Nemotron 3.5 Lightning is built for this high-volume execution, at a size that can run anywhere from an NVIDIA DGX Spark to the data center. See it running agentic…
- likes: 163
- replies: 5
- bookmarks: 28
- reposts: 6
NA NVIDIA AI@NVIDIAAINot every step in an agent workflow needs the same model. That’s why we’re also releasing NVIDIA NeMo Switchyard, a new open source library for model routing. Use frontier models for complex reasoning and planning, and Lightning for high-volume, specialized execution. Learn…
- likes: 143
- replies: 4
- bookmarks: 37
- reposts: 8
