← CanIRunLocal · All hardware · All models
OpenAI · 20.9B parameters (3.6B active) · chat, reasoning, coding.
Ollama weight files: MXFP4 14 GB (gpt-oss:20b).
Yes on 55 machines in this list. The smallest is RTX 3060 Ti 8GB: MXFP4, about 15.1 GB, roughly 14 tokens per second, via Ollama. ollama run gpt-oss:20b
ollama run gpt-oss:20bollama run gpt-oss:20bollama run gpt-oss:20bollama run gpt-oss:20blms get unsloth/gpt-oss-20b-GGUF@Q5_K_Mllama-server -hf unsloth/gpt-oss-20b-GGUF:Q5_K_M -c 8192vllm serve openai/gpt-oss-20b --max-model-len 8192 --gpu-memory-utilization 0.9| Hardware | Runtime | Quant | Memory | Fit | Speed | Command |
|---|---|---|---|---|---|---|
| RTX 3060 Ti 8GB | Ollama | MXFP4 | 15.1 GB | partial offload (slower) | ~14 tok/s | ollama run gpt-oss:20b |
| RTX 3070 8GB | Ollama | MXFP4 | 15.1 GB | partial offload (slower) | ~14 tok/s | ollama run gpt-oss:20b |
| RTX 4060 8GB | Ollama | MXFP4 | 15.1 GB | partial offload (slower) | ~13 tok/s | ollama run gpt-oss:20b |
| RX 7600 8GB | Ollama | MXFP4 | 15.1 GB | partial offload (slower) | ~13 tok/s | ollama run gpt-oss:20b |
| RTX 3080 10GB | Ollama | MXFP4 | 15.1 GB | partial offload (slower) | ~19 tok/s | ollama run gpt-oss:20b |
| RTX 3060 12GB | Ollama | MXFP4 | 15.1 GB | partial offload (slower) | ~22 tok/s | ollama run gpt-oss:20b |
| RTX 4070 12GB | Ollama | MXFP4 | 15.1 GB | partial offload (slower) | ~24 tok/s | ollama run gpt-oss:20b |
| RTX 5070 12GB | Ollama | MXFP4 | 15.1 GB | partial offload (slower) | ~26 tok/s | ollama run gpt-oss:20b |
| RTX 4060 Ti 16GB | Ollama | MXFP4 | 15.1 GB | fits in fast memory | ~36 tok/s | ollama run gpt-oss:20b |
| RTX 4070 Ti Super 16GB | Ollama | MXFP4 | 15.1 GB | fits in fast memory | ~84 tok/s | ollama run gpt-oss:20b |
| RTX 4080 16GB | Ollama | MXFP4 | 15.1 GB | fits in fast memory | ~89 tok/s | ollama run gpt-oss:20b |
| RTX 5080 16GB | Ollama | MXFP4 | 15.1 GB | fits in fast memory | ~119 tok/s | ollama run gpt-oss:20b |
| RX 7800 XT 16GB | Ollama | MXFP4 | 15.1 GB | fits in fast memory | ~78 tok/s | ollama run gpt-oss:20b |
| RX 9070 XT 16GB | Ollama | MXFP4 | 15.1 GB | fits in fast memory | ~80 tok/s | ollama run gpt-oss:20b |
| M2 24GB | Ollama | MXFP4 | 15.1 GB | fits in fast memory | ~17 tok/s | ollama run gpt-oss:20b |
| M3 24GB | Ollama | MXFP4 | 15.1 GB | fits in fast memory | ~17 tok/s | ollama run gpt-oss:20b |
| M4 24GB | Ollama | MXFP4 | 15.1 GB | fits in fast memory | ~20 tok/s | ollama run gpt-oss:20b |
| RTX 3090 24GB | Ollama | MXFP4 | 15.1 GB | fits in fast memory | ~116 tok/s | ollama run gpt-oss:20b |
| RTX 4090 24GB | Ollama | MXFP4 | 15.1 GB | fits in fast memory | ~125 tok/s | ollama run gpt-oss:20b |
| RX 7900 XTX 24GB | Ollama | MXFP4 | 15.1 GB | fits in fast memory | ~119 tok/s | ollama run gpt-oss:20b |
| M1 Pro 32GB | Ollama | MXFP4 | 15.1 GB | fits in fast memory | ~33 tok/s | ollama run gpt-oss:20b |
| M2 Pro 32GB | Ollama | MXFP4 | 15.1 GB | fits in fast memory | ~33 tok/s | ollama run gpt-oss:20b |
| M4 32GB | Ollama | MXFP4 | 15.1 GB | fits in fast memory | ~20 tok/s | ollama run gpt-oss:20b |
| M3 Pro 36GB | Ollama | MXFP4 | 15.1 GB | fits in fast memory | ~25 tok/s | ollama run gpt-oss:20b |
| Mac Studio M4 Max (32c GPU) 36GB | Ollama | MXFP4 | 15.1 GB | fits in fast memory | ~68 tok/s | ollama run gpt-oss:20b |
| Mac Studio M5 Max (32c GPU) 36GB | Ollama | MXFP4 | 15.1 GB | fits in fast memory | ~76 tok/s | ollama run gpt-oss:20b |
| RTX 5090 32GB | Ollama | MXFP4 | 15.1 GB | fits in fast memory | ~223 tok/s | ollama run gpt-oss:20b |
| M4 Pro 48GB | Ollama | MXFP4 | 15.1 GB | fits in fast memory | ~45 tok/s | ollama run gpt-oss:20b |
| Mac Studio M4 Max (40c GPU) 48GB | Ollama | MXFP4 | 15.1 GB | fits in fast memory | ~91 tok/s | ollama run gpt-oss:20b |
| Mac Studio M5 Max (40c GPU) 48GB | Ollama | MXFP4 | 15.1 GB | fits in fast memory | ~102 tok/s | ollama run gpt-oss:20b |
| M1 Max 64GB | Ollama | MXFP4 | 15.1 GB | fits in fast memory | ~66 tok/s | ollama run gpt-oss:20b |
| M2 Max 64GB | Ollama | MXFP4 | 15.1 GB | fits in fast memory | ~66 tok/s | ollama run gpt-oss:20b |
| M3 Max 64GB | Ollama | MXFP4 | 15.1 GB | fits in fast memory | ~66 tok/s | ollama run gpt-oss:20b |
| M4 Max 64GB | Ollama | MXFP4 | 15.1 GB | fits in fast memory | ~91 tok/s | ollama run gpt-oss:20b |
| Mac Studio M4 Max (40c GPU) 64GB | Ollama | MXFP4 | 15.1 GB | fits in fast memory | ~91 tok/s | ollama run gpt-oss:20b |
| Mac Studio M5 Max (40c GPU) 64GB | Ollama | MXFP4 | 15.1 GB | fits in fast memory | ~102 tok/s | ollama run gpt-oss:20b |
| DGX Spark 64GB (ASUS, Dell, HP, MSI, Acer, Gigabyte) | Ollama | MXFP4 | 15.1 GB | fits in fast memory | ~65 tok/s | ollama run gpt-oss:20b |
| Mac Studio M3 Ultra 96GB | Ollama | MXFP4 | 15.1 GB | fits in fast memory | ~136 tok/s | ollama run gpt-oss:20b |
| Mac Studio M5 Ultra 96GB | Ollama | MXFP4 | 15.1 GB | fits in fast memory | ~199 tok/s | ollama run gpt-oss:20b |
| M3 Max 128GB | Ollama | MXFP4 | 15.1 GB | fits in fast memory | ~66 tok/s | ollama run gpt-oss:20b |
| M4 Max 128GB | Ollama | MXFP4 | 15.1 GB | fits in fast memory | ~91 tok/s | ollama run gpt-oss:20b |
| Mac Studio M4 Max (40c GPU) 128GB | Ollama | MXFP4 | 15.1 GB | fits in fast memory | ~91 tok/s | ollama run gpt-oss:20b |
| Mac Studio M5 Max (40c GPU) 128GB | Ollama | MXFP4 | 15.1 GB | fits in fast memory | ~102 tok/s | ollama run gpt-oss:20b |
| DGX Spark 128GB (NVIDIA, ASUS, Dell, Lenovo, HP, MSI, Acer, Gigabyte) | Ollama | MXFP4 | 15.1 GB | fits in fast memory | ~65 tok/s | ollama run gpt-oss:20b |
| M2 Ultra 192GB | Ollama | MXFP4 | 15.1 GB | fits in fast memory | ~133 tok/s | ollama run gpt-oss:20b |
| 2x Mac Studio M5 Max (40c GPU) 128GB | llama.cpp | Q5_K_M GGUF | 12.8 GB | fits in fast memory | ~103 tok/s | llama-server -hf unsloth/gpt-oss-20b-GGUF:Q5_K_M --rpc node2:50052 -c 8192 # run rpc-server on each other node |
| Mac Studio M3 Ultra 256GB | Ollama | MXFP4 | 15.1 GB | fits in fast memory | ~136 tok/s | ollama run gpt-oss:20b |
| Mac Studio M5 Ultra (64c GPU) 256GB | Ollama | MXFP4 | 15.1 GB | fits in fast memory | ~199 tok/s | ollama run gpt-oss:20b |
| Mac Studio M5 Ultra (80c GPU) 256GB | Ollama | MXFP4 | 15.1 GB | fits in fast memory | ~199 tok/s | ollama run gpt-oss:20b |
| 2x DGX Spark 128GB | llama.cpp | Q5_K_M GGUF | 12.8 GB | fits in fast memory | ~66 tok/s | llama-server -hf unsloth/gpt-oss-20b-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 | 12.8 GB | fits in fast memory | ~91 tok/s | llama-server -hf unsloth/gpt-oss-20b-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 | 12.8 GB | fits in fast memory | ~202 tok/s | llama-server -hf unsloth/gpt-oss-20b-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 | 15.1 GB | fits in fast memory | ~136 tok/s | ollama run gpt-oss:20b |
| 4x DGX Spark 128GB | llama.cpp | Q5_K_M GGUF | 12.8 GB | fits in fast memory | ~58 tok/s | llama-server -hf unsloth/gpt-oss-20b-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 | 12.8 GB | fits in fast memory | ~178 tok/s | llama-server -hf unsloth/gpt-oss-20b-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