← CanIRunLocal · All hardware · All models
OpenAI · 117B parameters (5.1B active) · chat, reasoning, coding.
Ollama weight files: MXFP4 65 GB (gpt-oss:120b).
Yes on 18 machines in this list. The smallest is Mac Studio M3 Ultra 96GB: MXFP4, about 67.3 GB, roughly 116 tokens per second, via Ollama. ollama run gpt-oss:120b
ollama run gpt-oss:120bollama run gpt-oss:120bollama run gpt-oss:120bollama run gpt-oss:120blms get unsloth/gpt-oss-120b-GGUF@Q5_K_Mllama-server -hf unsloth/gpt-oss-120b-GGUF:Q5_K_M -c 8192vllm serve openai/gpt-oss-120b --max-model-len 8192 --gpu-memory-utilization 0.9| Hardware | Runtime | Quant | Memory | Fit | Speed | Command |
|---|---|---|---|---|---|---|
| Mac Studio M3 Ultra 96GB | Ollama | MXFP4 | 67.3 GB | fits in fast memory | ~116 tok/s | ollama run gpt-oss:120b |
| Mac Studio M5 Ultra 96GB | Ollama | MXFP4 | 67.3 GB | fits in fast memory | ~169 tok/s | ollama run gpt-oss:120b |
| M3 Max 128GB | Ollama | MXFP4 | 67.3 GB | fits in fast memory | ~56 tok/s | ollama run gpt-oss:120b |
| M4 Max 128GB | Ollama | MXFP4 | 67.3 GB | fits in fast memory | ~77 tok/s | ollama run gpt-oss:120b |
| Mac Studio M4 Max (40c GPU) 128GB | Ollama | MXFP4 | 67.3 GB | fits in fast memory | ~77 tok/s | ollama run gpt-oss:120b |
| Mac Studio M5 Max (40c GPU) 128GB | Ollama | MXFP4 | 67.3 GB | fits in fast memory | ~87 tok/s | ollama run gpt-oss:120b |
| DGX Spark 128GB (NVIDIA, ASUS, Dell, Lenovo, HP, MSI, Acer, Gigabyte) | Ollama | MXFP4 | 67.3 GB | fits in fast memory | ~55 tok/s | ollama run gpt-oss:120b |
| M2 Ultra 192GB | Ollama | MXFP4 | 67.3 GB | fits in fast memory | ~113 tok/s | ollama run gpt-oss:120b |
| 2x Mac Studio M5 Max (40c GPU) 128GB | llama.cpp | Q5_K_M GGUF | 65.2 GB | fits in fast memory | ~76 tok/s | llama-server -hf unsloth/gpt-oss-120b-GGUF:Q5_K_M --rpc node2:50052 -c 8192 # run rpc-server on each other node |
| Mac Studio M3 Ultra 256GB | Ollama | MXFP4 | 67.3 GB | fits in fast memory | ~116 tok/s | ollama run gpt-oss:120b |
| Mac Studio M5 Ultra (64c GPU) 256GB | Ollama | MXFP4 | 67.3 GB | fits in fast memory | ~169 tok/s | ollama run gpt-oss:120b |
| Mac Studio M5 Ultra (80c GPU) 256GB | Ollama | MXFP4 | 67.3 GB | fits in fast memory | ~169 tok/s | ollama run gpt-oss:120b |
| 2x DGX Spark 128GB | llama.cpp | Q5_K_M GGUF | 65.2 GB | fits in fast memory | ~49 tok/s | llama-server -hf unsloth/gpt-oss-120b-GGUF:Q5_K_M --rpc node2:50052 -c 8192 # run rpc-server on each other node |
| 4x Mac Studio M5 Max (40c GPU) 128GB | llama.cpp | Q5_K_M GGUF | 65.2 GB | fits in fast memory | ~67 tok/s | llama-server -hf unsloth/gpt-oss-120b-GGUF:Q5_K_M --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 | Q5_K_M GGUF | 65.2 GB | fits in fast memory | ~149 tok/s | llama-server -hf unsloth/gpt-oss-120b-GGUF:Q5_K_M --rpc node2:50052 -c 8192 # run rpc-server on each other node |
| Mac Studio M3 Ultra (80c GPU) 512GB | Ollama | MXFP4 | 67.3 GB | fits in fast memory | ~116 tok/s | ollama run gpt-oss:120b |
| 4x DGX Spark 128GB | llama.cpp | Q5_K_M GGUF | 65.2 GB | fits in fast memory | ~43 tok/s | llama-server -hf unsloth/gpt-oss-120b-GGUF:Q5_K_M --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 | Q5_K_M GGUF | 65.2 GB | fits in fast memory | ~131 tok/s | llama-server -hf unsloth/gpt-oss-120b-GGUF:Q5_K_M --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