How much VRAM does Nemotron 3.5 Lightning 30B A3B need?
Nemotron 3.5 Lightning 30B A3B needs 17.8 GB for its weights at 4-bit, 18.7 GB with an 8K context and 18.8 GB with 32K. At 8-bit, 32.1 GB.
- Parameters
- 31.6B
- Active per token
- 3B
- Longest context
- 256K
- Released
- Aug 2026
- License
- OpenMDW 1.1
- Strongest at
- Agents, Docs
Memory at each precision
| Precision | Weights | With 8K | With 32K | With 128K |
|---|---|---|---|---|
| 16-bit | 58.8 GB | 59.7 GB | 59.8 GB | 60.4 GB |
| 8-bit | 31.2 GB | 32.1 GB | 32.3 GB | 32.8 GB |
| 6-bit | 24.1 GB | 25.0 GB | 25.1 GB | 25.7 GB |
| 5-bit | 20.9 GB | 21.8 GB | 21.9 GB | 22.5 GB |
| 4-bit | 17.8 GB | 18.7 GB | 18.8 GB | 19.4 GB |
| 3-bit | 14.4 GB | 15.2 GB | 15.4 GB | 16.0 GB |
| 2-bit | 11.0 GB | 11.9 GB | 12.0 GB | 12.6 GB |
| 1.6-bit | 6.1 GB | 6.9 GB | 7.1 GB | 7.6 GB |
One conversation, 16-bit context cache, runtime overhead included.
What runs it
| Machine | Best fit | Context | Speed | First reply |
|---|---|---|---|---|
| RTX 3060 12 GB | Too big | |||
| RTX 4060 Ti 16 GB | Too big | |||
| RTX 3090 | 5-bit | 176K | 155 tok/s | 0.5 s |
| RTX 4090 | 5-bit | 176K | 161 tok/s | 0.2 s |
| RTX 5090 | 6-bit | 256K | 195 tok/s | 0.2 s |
| 2× RTX 3090 | 8-bit | 256K | 124 tok/s | 0.3 s |
| RTX PRO 6000 Blackwell | 8-bit | 256K | 175 tok/s | 0.1 s |
| M4 Max 128 GB | 8-bit | 256K | 61 tok/s | 2.9 s |
| M5 Max 128 GB | 8-bit | 256K | 65 tok/s | 0.9 s |
| Ryzen AI Max+ 395 128 GB | 8-bit | 256K | 39 tok/s | 1.7 s |
| DGX Spark 128 GB | 8-bit | 256K | 50 tok/s | 0.7 s |
| M3 Ultra 256 GB | 8-bit | 256K | 75 tok/s | 1.7 s |
| M5 Ultra 512 GB | 8-bit | 256K | 87 tok/s | 0.5 s |
| H100 SXM | 8-bit | 256K | 222 tok/s | 0.1 s |
| DGX Station GB300 | 8-bit | 256K | 265 tok/s | 0.1 s |
| 8× H200 NVL | 8-bit | 256K | 245 tok/s | 0.1 s |
Best fit is the highest quality, up to 8-bit, that leaves 16K of context. First reply is the wait for a 2K prompt. Speeds are rough, plus or minus 25%.
How to run it
- Ollama
ollama run nemotron-3.5-lightning:30b-a3b- llama.cpp
llama-server -hf unsloth/NVIDIA-Nemotron-3.5-Lightning-30B-A3B-GGUF- vLLM
vllm serve nvidia/NVIDIA-Nemotron-3.5-Lightning-30B-A3B-BF16- MLX (Mac)
mlx_lm.chat --model mlx-community/NVIDIA-Nemotron-3.5-Lightning-30B-A3B-4bit
Questions
How much VRAM does Nemotron 3.5 Lightning 30B A3B need?
Nemotron 3.5 Lightning 30B A3B needs 17.8 GB for its weights at 4-bit, 18.7 GB with an 8K context and 18.8 GB with 32K. At 8-bit, 32.1 GB.
Can Nemotron 3.5 Lightning 30B A3B run on an RTX 4090?
Yes. At 5-bit it holds 176K of context and writes about 161 tokens a second.