LocalClaw/LLM/Nemotron 3.5 Lightning 30B-A3B
Open-weight MoE

Nemotron 3.5 Lightning 30B-A3B

Catalogue summary: NVIDIA OpenMDW-1.1 hybrid Mamba/MoE/attention model for local agentic inference. The official GGUF path from ggml-org includes a 18.9GB Q4_0 build plus Ollama, llama.cpp and LM Studio recipes, with local contexts scaling from 4K to 256K+ depending on VRAM.

Repository editorial metadata; verify comparative claims in the linked upstream material.

48 GB catalogue minimum48 GB RAMQ4_0Coding assistant
Choose an app

Pick the app you already use. No terminal and no command to copy.

Compare models
Open in LM StudioOpens the app on this model's download screenOpen in UnslothLaunches Unsloth Desktop on this model's download pageOpen Ollama pageOpens the verified Ollama model pageOpen on Hugging FaceFiles, licence and available downloadsOpen llama.cpp setupAdvanced option: opens the model filesUse with LocalClawOptional workspace after the model is installed

Desktop app links require the app to be installed. If nothing opens, LocalClaw will show app-download and model-file fallbacks.

Parameters
30B (3B active, MoE)
Minimum RAM
48 GB
Model size
18.9 GB
Quantization
Q4_0

Can Nemotron 3.5 Lightning 30B-A3B run locally?

Nemotron 3.5 Lightning 30B-A3B has a catalogue minimum of 48 GB RAM with Q4_0. Actual memory use and speed vary by context length, runtime, backend and system headroom.

Use nvidia-nemotron-3.5-lightning-30b-a3b as the catalogue search term in a compatible runtime, and confirm the available format on the upstream repository before download.

Model sourcePublic GGUF repository (verified 2026-09-07): ggml-org/NVIDIA-Nemotron-3.5-Lightning-30B-A3B-GGUFUpstream releasePaper or model cardLicence text
chatcodereasoningagentictool-callinglong-contextpowergeneral

Install path

01
Check RAM fitMinimum 48 GB RAM. Start with the Q4_0 quant.
02
Load the modelSearch nvidia-nemotron-3.5-lightning-30b-a3b in LM Studio.
03
Control locallyUse LocalClaw to manage models, agents, chat, channels and scheduled OpenClaw work.

Catalogue record

  • Family: nemotron
  • Parameters: 30B (3B active, MoE)
  • Recommended quantization: Q4_0
  • Catalogue minimum RAM: 48 GB
  • Catalogue model size: 18.9 GB
  • Tags: chat, code, reasoning, agentic, tool-calling, long-context, power, general

Practical limits

  • Catalogue RAM is a minimum estimate, not a guarantee for every context length or runtime.
  • Speed and memory use vary by quantization, backend, context length and system headroom.
  • Verify architecture, licence and usage restrictions in the linked upstream material before deployment.

Catalogue tags

  • chat
  • code
  • reasoning
  • agentic
  • tool-calling
  • long-context

Capability profile

Repository catalogue ratings used by LocalClaw's editorial rubric. They are not a standardized third-party benchmark.

speed
7
quality
8
coding
8
reasoning
8

Technical notes

Developer
NVIDIA
License
OpenMDW-1.1
Context window
1,048,576 tokens
Architecture
Hybrid 30B mixture-of-experts model with about 3B active parameters per token, interleaved Mamba-2 layers, MoE layers, selected attention layers and Multi-Token Prediction support. NVIDIA publishes BF16 and NVFP4 checkpoints, while ggml-org provides the official GGUF path for llama.cpp-compatible local runtimes.

This model fits these next steps

Hardware fit is based on LocalClaw's RAM tier, model size and quantization metadata. Always leave memory headroom for your OS and runtime.

Workstation fitMac Studio M4 Max 64GBCUDA workstation classNVIDIA GB10 / DGX SparkLarge local models64GB RAM guide

Related catalogue entries

Linked mechanically by family, shared tags and nearby RAM tier; this is not a quality ranking.

Llama-3.1-Nemotron (70B) 70BLlama-3.3-Nemotron-Super (49B) 49BApriel Nemotron 15B Thinker 15BNemotron Nano 9B v2 9BNemotron Mini (4B) 4BOrnith-1.5-35B-A3B 35B (3B active, MoE)KAT-Coder V2.5 Dev 35B (3B active, MoE)K2-Horizon-MoVA-36B-A4B 36B (4B active, MoE)

Where to go next