← CanIRunLocal · All hardware · All models
NVIDIA · 31.6B parameters (3B active) · chat, reasoning, coding.
Ollama weight files: Q4_K_M 25 GB (nemotron-3.5-lightning:30b-a3b-q4_K_M); Q8_0 35 GB (nemotron-3.5-lightning:30b-a3b-q8_0).
Yes on 35 machines in this list. The smallest is RTX 3090 24GB: Q4_K_M, about 26.4 GB, roughly 49 tokens per second, via Ollama. ollama run nemotron-3.5-lightning:30b-a3b-q4_K_M
ollama run nemotron-3.5-lightning:30b-a3b-q8_0ollama run nemotron-3.5-lightning:30b-a3b-q8_0ollama run nemotron-3.5-lightning:30b-a3b-q8_0ollama run nemotron-3.5-lightning:30b-a3b-q8_0lms get unsloth/NVIDIA-Nemotron-3.5-Lightning-30B-A3B-GGUF@Q8_0llama-server -hf unsloth/NVIDIA-Nemotron-3.5-Lightning-30B-A3B-GGUF:Q8_0 -c 8192vllm serve nvidia/NVIDIA-Nemotron-3.5-Lightning-30B-A3B-BF16 --max-model-len 8192 --gpu-memory-utilization 0.9| Hardware | Runtime | Quant | Memory | Fit | Speed | Command |
|---|---|---|---|---|---|---|
| RTX 3090 24GB | Ollama | Q4_K_M | 26.4 GB | partial offload (slower) | ~49 tok/s | ollama run nemotron-3.5-lightning:30b-a3b-q4_K_M |
| RTX 4090 24GB | Ollama | Q4_K_M | 26.4 GB | partial offload (slower) | ~50 tok/s | ollama run nemotron-3.5-lightning:30b-a3b-q4_K_M |
| RX 7900 XTX 24GB | Ollama | Q4_K_M | 26.4 GB | partial offload (slower) | ~49 tok/s | ollama run nemotron-3.5-lightning:30b-a3b-q4_K_M |
| M3 Pro 36GB | Ollama | Q4_K_M | 26.4 GB | fits in fast memory | ~25 tok/s | ollama run nemotron-3.5-lightning:30b-a3b-q4_K_M |
| Mac Studio M4 Max (32c GPU) 36GB | Ollama | Q4_K_M | 26.4 GB | fits in fast memory | ~69 tok/s | ollama run nemotron-3.5-lightning:30b-a3b-q4_K_M |
| Mac Studio M5 Max (32c GPU) 36GB | Ollama | Q4_K_M | 26.4 GB | fits in fast memory | ~78 tok/s | ollama run nemotron-3.5-lightning:30b-a3b-q4_K_M |
| RTX 5090 32GB | Ollama | Q4_K_M | 26.4 GB | fits in fast memory | ~227 tok/s | ollama run nemotron-3.5-lightning:30b-a3b-q4_K_M |
| M4 Pro 48GB | Ollama | Q4_K_M | 26.4 GB | fits in fast memory | ~46 tok/s | ollama run nemotron-3.5-lightning:30b-a3b-q4_K_M |
| Mac Studio M4 Max (40c GPU) 48GB | Ollama | Q4_K_M | 26.4 GB | fits in fast memory | ~92 tok/s | ollama run nemotron-3.5-lightning:30b-a3b-q4_K_M |
| Mac Studio M5 Max (40c GPU) 48GB | Ollama | Q4_K_M | 26.4 GB | fits in fast memory | ~103 tok/s | ollama run nemotron-3.5-lightning:30b-a3b-q4_K_M |
| M1 Max 64GB | Ollama | Q8_0 | 36.4 GB | fits in fast memory | ~48 tok/s | ollama run nemotron-3.5-lightning:30b-a3b-q8_0 |
| M2 Max 64GB | Ollama | Q8_0 | 36.4 GB | fits in fast memory | ~48 tok/s | ollama run nemotron-3.5-lightning:30b-a3b-q8_0 |
| M3 Max 64GB | Ollama | Q8_0 | 36.4 GB | fits in fast memory | ~48 tok/s | ollama run nemotron-3.5-lightning:30b-a3b-q8_0 |
| M4 Max 64GB | Ollama | Q8_0 | 36.4 GB | fits in fast memory | ~66 tok/s | ollama run nemotron-3.5-lightning:30b-a3b-q8_0 |
| Mac Studio M4 Max (40c GPU) 64GB | Ollama | Q8_0 | 36.4 GB | fits in fast memory | ~66 tok/s | ollama run nemotron-3.5-lightning:30b-a3b-q8_0 |
| Mac Studio M5 Max (40c GPU) 64GB | Ollama | Q8_0 | 36.4 GB | fits in fast memory | ~74 tok/s | ollama run nemotron-3.5-lightning:30b-a3b-q8_0 |
| DGX Spark 64GB (ASUS, Dell, HP, MSI, Acer, Gigabyte) | Ollama | Q8_0 | 36.4 GB | fits in fast memory | ~47 tok/s | ollama run nemotron-3.5-lightning:30b-a3b-q8_0 |
| Mac Studio M3 Ultra 96GB | Ollama | Q8_0 | 36.4 GB | fits in fast memory | ~99 tok/s | ollama run nemotron-3.5-lightning:30b-a3b-q8_0 |
| Mac Studio M5 Ultra 96GB | Ollama | Q8_0 | 36.4 GB | fits in fast memory | ~144 tok/s | ollama run nemotron-3.5-lightning:30b-a3b-q8_0 |
| M3 Max 128GB | Ollama | Q8_0 | 36.4 GB | fits in fast memory | ~48 tok/s | ollama run nemotron-3.5-lightning:30b-a3b-q8_0 |
| M4 Max 128GB | Ollama | Q8_0 | 36.4 GB | fits in fast memory | ~66 tok/s | ollama run nemotron-3.5-lightning:30b-a3b-q8_0 |
| Mac Studio M4 Max (40c GPU) 128GB | Ollama | Q8_0 | 36.4 GB | fits in fast memory | ~66 tok/s | ollama run nemotron-3.5-lightning:30b-a3b-q8_0 |
| Mac Studio M5 Max (40c GPU) 128GB | Ollama | Q8_0 | 36.4 GB | fits in fast memory | ~74 tok/s | ollama run nemotron-3.5-lightning:30b-a3b-q8_0 |
| DGX Spark 128GB (NVIDIA, ASUS, Dell, Lenovo, HP, MSI, Acer, Gigabyte) | Ollama | Q8_0 | 36.4 GB | fits in fast memory | ~47 tok/s | ollama run nemotron-3.5-lightning:30b-a3b-q8_0 |
| M2 Ultra 192GB | Ollama | Q8_0 | 36.4 GB | fits in fast memory | ~96 tok/s | ollama run nemotron-3.5-lightning:30b-a3b-q8_0 |
| 2x Mac Studio M5 Max (40c GPU) 128GB | llama.cpp | Q8_0 GGUF | 36.4 GB | fits in fast memory | ~63 tok/s | llama-server -hf unsloth/NVIDIA-Nemotron-3.5-Lightning-30B-A3B-GGUF:Q8_0 --rpc node2:50052 -c 8192 # run rpc-server on each other node |
| Mac Studio M3 Ultra 256GB | Ollama | Q8_0 | 36.4 GB | fits in fast memory | ~99 tok/s | ollama run nemotron-3.5-lightning:30b-a3b-q8_0 |
| Mac Studio M5 Ultra (64c GPU) 256GB | Ollama | Q8_0 | 36.4 GB | fits in fast memory | ~144 tok/s | ollama run nemotron-3.5-lightning:30b-a3b-q8_0 |
| Mac Studio M5 Ultra (80c GPU) 256GB | Ollama | Q8_0 | 36.4 GB | fits in fast memory | ~144 tok/s | ollama run nemotron-3.5-lightning:30b-a3b-q8_0 |
| 2x DGX Spark 128GB | llama.cpp | Q8_0 GGUF | 36.4 GB | fits in fast memory | ~40 tok/s | llama-server -hf unsloth/NVIDIA-Nemotron-3.5-Lightning-30B-A3B-GGUF:Q8_0 --rpc node2:50052 -c 8192 # run rpc-server on each other node |
| 4x Mac Studio M5 Max (40c GPU) 128GB | llama.cpp | Q8_0 GGUF | 36.4 GB | fits in fast memory | ~55 tok/s | llama-server -hf unsloth/NVIDIA-Nemotron-3.5-Lightning-30B-A3B-GGUF:Q8_0 --rpc node2:50052,node3:50052,node4:50052 -c 8192 # run rpc-server on each other node |
| 2x Mac Studio M5 Ultra (80c GPU) 256GB | llama.cpp | Q8_0 GGUF | 36.4 GB | fits in fast memory | ~123 tok/s | llama-server -hf unsloth/NVIDIA-Nemotron-3.5-Lightning-30B-A3B-GGUF:Q8_0 --rpc node2:50052 -c 8192 # run rpc-server on each other node |
| Mac Studio M3 Ultra (80c GPU) 512GB | Ollama | Q8_0 | 36.4 GB | fits in fast memory | ~99 tok/s | ollama run nemotron-3.5-lightning:30b-a3b-q8_0 |
| 4x DGX Spark 128GB | llama.cpp | Q8_0 GGUF | 36.4 GB | fits in fast memory | ~35 tok/s | llama-server -hf unsloth/NVIDIA-Nemotron-3.5-Lightning-30B-A3B-GGUF:Q8_0 --rpc node2:50052,node3:50052,node4:50052 -c 8192 # run rpc-server on each other node |
| 4x Mac Studio M5 Ultra (80c GPU) 256GB | llama.cpp | Q8_0 GGUF | 36.4 GB | fits in fast memory | ~108 tok/s | llama-server -hf unsloth/NVIDIA-Nemotron-3.5-Lightning-30B-A3B-GGUF:Q8_0 --rpc node2:50052,node3:50052,node4:50052 -c 8192 # run rpc-server on each other node |
Estimates, not benchmarks. Model sizes are ollama.com download sizes; memory needed adds headroom for ~8k tokens of context; speeds are rough figures from memory bandwidth. Data updated 2026-10-18.
A project by OnDevice Labs · llms.txt