Written by Jakub Rusinowski · Last updated September 19, 2026
Model library → Nemotron 3.5 → Nemotron 3.5 Lightning 30B-A3B
30B total, 3B active, pre-trained on over 20T tokens and released with speculative-decoding companions. Hybrid Mamba-2 + MoE + attention, so its cache does not grow the way a pure transformer's would. BF16 and NVFP4 builds are published.
Nemotron 3.5 Lightning 30B-A3B needs about 19 GB of VRAM at Q4_K_M — quantized weights plus framework overhead, before any KV cache. On Apple Silicon that figure comes out of unified memory.
| Parameters | 30 Billion (3B active) |
| Context window | 1,000,000 |
| Architecture | Hybrid Mamba-2 + MoE |
| Provider | NVIDIA |
| Licence | OpenMDW-1.1 |
| Specified at | Q4_K_M |
| System RAM | 32 GB |
| Record updated | 2026-09-19 |
Corroborated — Two or more independent sources agree on these figures, but the model card itself was not retrieved. Treat the numbers as good rather than confirmed. Still unconfirmed: arch.kv_layers.
OpenMDW-1.1 — commercial use permitted. Commercial use permitted. No usage restrictions beyond attribution.
Modelled on a reference NVIDIA RTX 4090 (24 GB). Weights plus framework overhead only — this model publishes no architecture we can read, so no KV cache is included. A real session needs more; the figure is a floor, not a target. Speed figures are ESTIMATES from the memory-bandwidth roofline described on the methodology page, not benchmarks we ran — rows marked measured come from published or reader-submitted runs. VRAM here includes the KV cache, so it reads higher than the headline figure above, which does not.
| Quant | Bits/weight | Weights | VRAM needed | Est. speed | Fit on 24 GB |
|---|---|---|---|---|---|
| Q2_K | 2.63 | 9.9 GB | 10.7 GB | ~212 tok/s (est.) | Fits comfortably |
| Q3_K_M | 3.41 | 12.8 GB | 13.6 GB | ~196 tok/s (est.) | Fits comfortably |
| Q4_K_M | 4.83 | 18.1 GB | 18.9 GB | ~172 tok/s (est.) | Fits comfortably |
| Q5_K_M | 5.67 | 21.3 GB | 22.1 GB | ~160 tok/s (est.) | Tight fit |
| Q6_K | 6.56 | 24.6 GB | 25.4 GB | ~22 tok/s (est.) | Offloads to system RAM (slow) |
| Q8_0 | 8.50 | 31.9 GB | 32.7 GB | ~20 tok/s (est.) | Offloads to system RAM (slow) |
| F16 | 16.00 | 60 GB | 60.8 GB | — | Won't fit |
Want to set your own context length and KV-cache quantization? Use the interactive VRAM calculator.
or compare on Vast.ai from $0.35/hr (typical low · varies)
As an Amazon Associate we earn from qualifying purchases. Cloud GPU links are referral links — we may earn a commission at no extra cost to you.
The cheapest catalogued GPU that runs Nemotron 3.5 Lightning 30B-A3B is the AMD Radeon RX 7900 XT (20 GB).
Install Ollama, then run:
ollama run nemotron-3.5-lightning:30bWeights on Hugging Face: nvidia/NVIDIA-Nemotron-3.5-Lightning-30B-A3B-BF16.
Best for: agents, long context, tool calling, reasoning.
← All Nemotron 3.5 models | VRAM calculator | Check your own hardware