← CanIRunLocal · All hardware · All models
Microsoft · 14.7B parameters · chat, reasoning.
Ollama weight files: Q4_K_M 9.1 GB (phi4:14b); Q8_0 16 GB (phi4:14b-q8_0).
Yes on 59 machines in this list. The smallest is RTX 3060 Ti 8GB: Q4_K_M, about 10.4 GB, roughly 11 tokens per second, via Ollama. ollama run phi4:14b
ollama run phi4:14b-q8_0ollama run phi4:14b-q8_0ollama run phi4:14b-q8_0ollama run phi4:14b-q8_0lms get unsloth/phi-4-GGUF@Q8_0llama-server -hf unsloth/phi-4-GGUF:Q8_0 -c 8192vllm serve microsoft/phi-4 --max-model-len 8192 --gpu-memory-utilization 0.9| Hardware | Runtime | Quant | Memory | Fit | Speed | Command |
|---|---|---|---|---|---|---|
| RTX 3060 Ti 8GB | Ollama | Q4_K_M | 10.4 GB | partial offload (slower) | ~11 tok/s | ollama run phi4:14b |
| RTX 3070 8GB | Ollama | Q4_K_M | 10.4 GB | partial offload (slower) | ~11 tok/s | ollama run phi4:14b |
| RTX 4060 8GB | Ollama | Q4_K_M | 10.4 GB | partial offload (slower) | ~10 tok/s | ollama run phi4:14b |
| RX 7600 8GB | Ollama | Q4_K_M | 10.4 GB | partial offload (slower) | ~10 tok/s | ollama run phi4:14b |
| RTX 3080 10GB | Ollama | Q4_K_M | 10.4 GB | partial offload (slower) | ~26 tok/s | ollama run phi4:14b |
| M1 16GB | Ollama | Q4_K_M | 10.4 GB | fits in fast memory | ~6 tok/s | ollama run phi4:14b |
| M2 16GB | Ollama | Q4_K_M | 10.4 GB | fits in fast memory | ~9 tok/s | ollama run phi4:14b |
| M3 16GB | Ollama | Q4_K_M | 10.4 GB | fits in fast memory | ~9 tok/s | ollama run phi4:14b |
| M4 16GB | Ollama | Q4_K_M | 10.4 GB | fits in fast memory | ~11 tok/s | ollama run phi4:14b |
| RTX 3060 12GB | Ollama | Q4_K_M | 10.4 GB | fits in fast memory | ~24 tok/s | ollama run phi4:14b |
| RTX 4070 12GB | Ollama | Q4_K_M | 10.4 GB | fits in fast memory | ~33 tok/s | ollama run phi4:14b |
| RTX 5070 12GB | Ollama | Q4_K_M | 10.4 GB | fits in fast memory | ~44 tok/s | ollama run phi4:14b |
| RTX 4060 Ti 16GB | Ollama | Q4_K_M | 10.4 GB | fits in fast memory | ~19 tok/s | ollama run phi4:14b |
| RTX 4070 Ti Super 16GB | Ollama | Q4_K_M | 10.4 GB | fits in fast memory | ~44 tok/s | ollama run phi4:14b |
| RTX 4080 16GB | Ollama | Q4_K_M | 10.4 GB | fits in fast memory | ~47 tok/s | ollama run phi4:14b |
| RTX 5080 16GB | Ollama | Q4_K_M | 10.4 GB | fits in fast memory | ~63 tok/s | ollama run phi4:14b |
| RX 7800 XT 16GB | Ollama | Q4_K_M | 10.4 GB | fits in fast memory | ~41 tok/s | ollama run phi4:14b |
| RX 9070 XT 16GB | Ollama | Q4_K_M | 10.4 GB | fits in fast memory | ~42 tok/s | ollama run phi4:14b |
| M2 24GB | Ollama | Q4_K_M | 10.4 GB | fits in fast memory | ~9 tok/s | ollama run phi4:14b |
| M3 24GB | Ollama | Q4_K_M | 10.4 GB | fits in fast memory | ~9 tok/s | ollama run phi4:14b |
| M4 24GB | Ollama | Q4_K_M | 10.4 GB | fits in fast memory | ~11 tok/s | ollama run phi4:14b |
| RTX 3090 24GB | Ollama | Q8_0 | 17.3 GB | fits in fast memory | ~35 tok/s | ollama run phi4:14b-q8_0 |
| RTX 4090 24GB | Ollama | Q8_0 | 17.3 GB | fits in fast memory | ~38 tok/s | ollama run phi4:14b-q8_0 |
| RX 7900 XTX 24GB | Ollama | Q8_0 | 17.3 GB | fits in fast memory | ~36 tok/s | ollama run phi4:14b-q8_0 |
| M1 Pro 32GB | Ollama | Q8_0 | 17.3 GB | fits in fast memory | ~10 tok/s | ollama run phi4:14b-q8_0 |
| M2 Pro 32GB | Ollama | Q8_0 | 17.3 GB | fits in fast memory | ~10 tok/s | ollama run phi4:14b-q8_0 |
| M4 32GB | Ollama | Q8_0 | 17.3 GB | fits in fast memory | ~6 tok/s | ollama run phi4:14b-q8_0 |
| M3 Pro 36GB | Ollama | Q8_0 | 17.3 GB | fits in fast memory | ~8 tok/s | ollama run phi4:14b-q8_0 |
| Mac Studio M4 Max (32c GPU) 36GB | Ollama | Q8_0 | 17.3 GB | fits in fast memory | ~21 tok/s | ollama run phi4:14b-q8_0 |
| Mac Studio M5 Max (32c GPU) 36GB | Ollama | Q8_0 | 17.3 GB | fits in fast memory | ~23 tok/s | ollama run phi4:14b-q8_0 |
| RTX 5090 32GB | Ollama | Q8_0 | 17.3 GB | fits in fast memory | ~67 tok/s | ollama run phi4:14b-q8_0 |
| M4 Pro 48GB | Ollama | Q8_0 | 17.3 GB | fits in fast memory | ~14 tok/s | ollama run phi4:14b-q8_0 |
| Mac Studio M4 Max (40c GPU) 48GB | Ollama | Q8_0 | 17.3 GB | fits in fast memory | ~27 tok/s | ollama run phi4:14b-q8_0 |
| Mac Studio M5 Max (40c GPU) 48GB | Ollama | Q8_0 | 17.3 GB | fits in fast memory | ~31 tok/s | ollama run phi4:14b-q8_0 |
| M1 Max 64GB | Ollama | Q8_0 | 17.3 GB | fits in fast memory | ~20 tok/s | ollama run phi4:14b-q8_0 |
| M2 Max 64GB | Ollama | Q8_0 | 17.3 GB | fits in fast memory | ~20 tok/s | ollama run phi4:14b-q8_0 |
| M3 Max 64GB | Ollama | Q8_0 | 17.3 GB | fits in fast memory | ~20 tok/s | ollama run phi4:14b-q8_0 |
| M4 Max 64GB | Ollama | Q8_0 | 17.3 GB | fits in fast memory | ~27 tok/s | ollama run phi4:14b-q8_0 |
| Mac Studio M4 Max (40c GPU) 64GB | Ollama | Q8_0 | 17.3 GB | fits in fast memory | ~27 tok/s | ollama run phi4:14b-q8_0 |
| Mac Studio M5 Max (40c GPU) 64GB | Ollama | Q8_0 | 17.3 GB | fits in fast memory | ~31 tok/s | ollama run phi4:14b-q8_0 |
| DGX Spark 64GB (ASUS, Dell, HP, MSI, Acer, Gigabyte) | Ollama | Q8_0 | 17.3 GB | fits in fast memory | ~13 tok/s | ollama run phi4:14b-q8_0 |
| Mac Studio M3 Ultra 96GB | Ollama | Q8_0 | 17.3 GB | fits in fast memory | ~41 tok/s | ollama run phi4:14b-q8_0 |
| Mac Studio M5 Ultra 96GB | Ollama | Q8_0 | 17.3 GB | fits in fast memory | ~60 tok/s | ollama run phi4:14b-q8_0 |
| M3 Max 128GB | Ollama | Q8_0 | 17.3 GB | fits in fast memory | ~20 tok/s | ollama run phi4:14b-q8_0 |
| M4 Max 128GB | Ollama | Q8_0 | 17.3 GB | fits in fast memory | ~27 tok/s | ollama run phi4:14b-q8_0 |
| Mac Studio M4 Max (40c GPU) 128GB | Ollama | Q8_0 | 17.3 GB | fits in fast memory | ~27 tok/s | ollama run phi4:14b-q8_0 |
| Mac Studio M5 Max (40c GPU) 128GB | Ollama | Q8_0 | 17.3 GB | fits in fast memory | ~31 tok/s | ollama run phi4:14b-q8_0 |
| DGX Spark 128GB (NVIDIA, ASUS, Dell, Lenovo, HP, MSI, Acer, Gigabyte) | Ollama | Q8_0 | 17.3 GB | fits in fast memory | ~13 tok/s | ollama run phi4:14b-q8_0 |
| M2 Ultra 192GB | Ollama | Q8_0 | 17.3 GB | fits in fast memory | ~40 tok/s | ollama run phi4:14b-q8_0 |
| 2x Mac Studio M5 Max (40c GPU) 128GB | llama.cpp | Q8_0 GGUF | 16.9 GB | fits in fast memory | ~27 tok/s | llama-server -hf unsloth/phi-4-GGUF:Q8_0 --rpc node2:50052 -c 8192 # run rpc-server on each other node |
| Mac Studio M3 Ultra 256GB | Ollama | Q8_0 | 17.3 GB | fits in fast memory | ~41 tok/s | ollama run phi4:14b-q8_0 |
| Mac Studio M5 Ultra (64c GPU) 256GB | Ollama | Q8_0 | 17.3 GB | fits in fast memory | ~60 tok/s | ollama run phi4:14b-q8_0 |
| Mac Studio M5 Ultra (80c GPU) 256GB | Ollama | Q8_0 | 17.3 GB | fits in fast memory | ~60 tok/s | ollama run phi4:14b-q8_0 |
| 2x DGX Spark 128GB | llama.cpp | Q8_0 GGUF | 16.9 GB | fits in fast memory | ~11 tok/s | llama-server -hf unsloth/phi-4-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 | 16.9 GB | fits in fast memory | ~24 tok/s | llama-server -hf unsloth/phi-4-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 | 16.9 GB | fits in fast memory | ~52 tok/s | llama-server -hf unsloth/phi-4-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 | 17.3 GB | fits in fast memory | ~41 tok/s | ollama run phi4:14b-q8_0 |
| 4x DGX Spark 128GB | llama.cpp | Q8_0 GGUF | 16.9 GB | fits in fast memory | ~10 tok/s | llama-server -hf unsloth/phi-4-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 | 16.9 GB | fits in fast memory | ~46 tok/s | llama-server -hf unsloth/phi-4-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