← CanIRunLocal · All hardware · All models
Meta · 109B parameters (17B active) · chat, vision.
Ollama weight files: Q4_K_M 67 GB (llama4:16x17b).
Yes on 18 machines in this list. The smallest is Mac Studio M3 Ultra 96GB: Q4_K_M, about 69.3 GB, roughly 31 tokens per second, via Ollama. ollama run llama4:16x17b
ollama run llama4:16x17bollama run llama4:16x17bollama run llama4:16x17bollama run llama4:16x17blms get unsloth/Llama-4-Scout-17B-16E-Instruct-GGUF@Q5_K_Mllama-server -hf unsloth/Llama-4-Scout-17B-16E-Instruct-GGUF:Q5_K_M -c 8192vllm serve RedHatAI/Llama-4-Scout-17B-16E-Instruct-quantized.w4a16 --max-model-len 8192 --gpu-memory-utilization 0.9| Hardware | Runtime | Quant | Memory | Fit | Speed | Command |
|---|---|---|---|---|---|---|
| Mac Studio M3 Ultra 96GB | Ollama | Q4_K_M | 69.3 GB | fits in fast memory | ~31 tok/s | ollama run llama4:16x17b |
| Mac Studio M5 Ultra 96GB | Ollama | Q4_K_M | 69.3 GB | fits in fast memory | ~46 tok/s | ollama run llama4:16x17b |
| M3 Max 128GB | Ollama | Q4_K_M | 69.3 GB | fits in fast memory | ~15 tok/s | ollama run llama4:16x17b |
| M4 Max 128GB | Ollama | Q4_K_M | 69.3 GB | fits in fast memory | ~21 tok/s | ollama run llama4:16x17b |
| Mac Studio M4 Max (40c GPU) 128GB | Ollama | Q4_K_M | 69.3 GB | fits in fast memory | ~21 tok/s | ollama run llama4:16x17b |
| Mac Studio M5 Max (40c GPU) 128GB | Ollama | Q4_K_M | 69.3 GB | fits in fast memory | ~24 tok/s | ollama run llama4:16x17b |
| DGX Spark 128GB (NVIDIA, ASUS, Dell, Lenovo, HP, MSI, Acer, Gigabyte) | Ollama | Q4_K_M | 69.3 GB | fits in fast memory | ~15 tok/s | ollama run llama4:16x17b |
| M2 Ultra 192GB | Ollama | Q4_K_M | 69.3 GB | fits in fast memory | ~31 tok/s | ollama run llama4:16x17b |
| 2x Mac Studio M5 Max (40c GPU) 128GB | llama.cpp | Q8_0 GGUF | 116.9 GB | fits in fast memory | ~12 tok/s | llama-server -hf unsloth/Llama-4-Scout-17B-16E-Instruct-GGUF:Q8_0 --rpc node2:50052 -c 8192 # run rpc-server on each other node |
| Mac Studio M3 Ultra 256GB | Ollama | Q4_K_M | 69.3 GB | fits in fast memory | ~31 tok/s | ollama run llama4:16x17b |
| Mac Studio M5 Ultra (64c GPU) 256GB | Ollama | Q4_K_M | 69.3 GB | fits in fast memory | ~46 tok/s | ollama run llama4:16x17b |
| Mac Studio M5 Ultra (80c GPU) 256GB | Ollama | Q4_K_M | 69.3 GB | fits in fast memory | ~46 tok/s | ollama run llama4:16x17b |
| 2x DGX Spark 128GB | llama.cpp | Q8_0 GGUF | 116.9 GB | fits in fast memory | ~7 tok/s | llama-server -hf unsloth/Llama-4-Scout-17B-16E-Instruct-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 | 116.9 GB | fits in fast memory | ~10 tok/s | llama-server -hf unsloth/Llama-4-Scout-17B-16E-Instruct-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 | 116.9 GB | fits in fast memory | ~23 tok/s | llama-server -hf unsloth/Llama-4-Scout-17B-16E-Instruct-GGUF:Q8_0 --rpc node2:50052 -c 8192 # run rpc-server on each other node |
| Mac Studio M3 Ultra (80c GPU) 512GB | Ollama | Q4_K_M | 69.3 GB | fits in fast memory | ~31 tok/s | ollama run llama4:16x17b |
| 4x DGX Spark 128GB | llama.cpp | Q8_0 GGUF | 116.9 GB | fits in fast memory | ~7 tok/s | llama-server -hf unsloth/Llama-4-Scout-17B-16E-Instruct-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 | 116.9 GB | fits in fast memory | ~20 tok/s | llama-server -hf unsloth/Llama-4-Scout-17B-16E-Instruct-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 2026-10-16.
A project by OnDevice Labs · llms.txt