Can You Run Nemotron 3.5 Lightning (30B-A3B)? VRAM Requirements
Nemotron 3.5 Lightning (30B-A3B) has 30B total parameters (MoE, 3B active per token). Running it at NVFP4 with 16K context takes roughly 16.4 GB of memory (14.8 GB weights + 0.1 GB KV cache + runtime overhead). The minimum viable hardware is about RTX 3090 / 4090 24GB. All numbers are planning estimates, verified 2026-07-27.
VRAM by quantization (16K context)
| Quantization | Weights | Total needed | Minimum hardware |
|---|---|---|---|
| FP16 / BF16 | 55.9 GB | 60.4 GB | A100 / H100 80GB |
| FP8 | 27.9 GB | 30.3 GB | Mac 36GB unified |
| Q8_0 (GGUF) | 29.6 GB | 32.1 GB | Mac 36GB unified |
| Q6_K (GGUF) | 22.9 GB | 24.8 GB | RTX 5090 32GB |
| Q5_K_M (GGUF) | 19.8 GB | 21.5 GB | RTX 3090 / 4090 24GB |
| MXFP4 | 14.8 GB | 16.4 GB | RTX 3090 / 4090 24GB |
| NVFP4recommended | 14.8 GB | 16.4 GB | RTX 3090 / 4090 24GB |
| Q4_K_M (GGUF) | 17.0 GB | 18.6 GB | RTX 3090 / 4090 24GB |
| Q3_K_M (GGUF) | 13.1 GB | 14.7 GB | RTX 3090 / 4090 24GB |
| Q2_K (GGUF) | 9.8 GB | 11.4 GB | RTX 4060 Ti 16GB |
How context length changes memory (NVFP4)
| Context | KV cache | Total needed |
|---|---|---|
| 4K | 0.0 GB | 16.3 GB |
| 16K | 0.1 GB | 16.4 GB |
| 64K | 0.4 GB | 16.7 GB |
| 128K | 0.8 GB | 17.1 GB |
| 256K | 1.5 GB | 17.8 GB |
KV cache grows linearly with context; FP8 KV cache halves it again (try it in the calculator below).
Customize the estimate
LLM VRAM Calculator — Can I Run It?
Estimate how much GPU memory (VRAM) you need to run open-weight LLMs like Kimi K3, DeepSeek R1, Qwen3, or Llama locally. Pick a model, quantization, and context length — the calculator adds up weights, KV cache, and runtime overhead, then shows which GPUs or Macs can fit it. Runs entirely in your browser: no upload, no signup.
30B total params · 3B active (MoE) · max 256K context
NVIDIA 4-bit float, Nemotron native · ~0.53 bytes/param
Estimated memory needed
17.1 GB
Will it fit?
Assumes ~92% of device memory is usable. Multi-GPU counts are for tensor/pipeline parallel serving (vLLM, SGLang); Apple Silicon uses unified memory via llama.cpp or MLX.
FAQ
How much VRAM does Nemotron 3.5 Lightning (30B-A3B) need?
About 16.4 GB at the recommended NVFP4 quantization with 16K context: 14.8 GB for weights, 0.1 GB for KV cache, plus runtime overhead. Longer context grows the KV cache.
What is the minimum hardware for Nemotron 3.5 Lightning (30B-A3B)?
Roughly RTX 3090 / 4090 24GB, assuming 92% of device memory is usable. Lower-end setups can try more aggressive quantization (Q3/Q2) at a noticeable quality cost.
Can I run Nemotron 3.5 Lightning (30B-A3B) on a Mac?
Yes. At NVFP4 with 16K context it needs about 16.4 GB, which fits the unified memory of a Mac M3 Ultra 512GB via llama.cpp or MLX.
Which quantization should I use for Nemotron 3.5 Lightning (30B-A3B)?
NVFP4 is the model's native/recommended format — NVIDIA 4-bit float, Nemotron native. Pick Q4_K_M for a smaller footprint or FP8 / Q8_0 for near-lossless quality.