← CanIRunLocal · All hardware · All models
Alibaba · 0.75B parameters · chat, reasoning.
Ollama weight files: Q4_K_M 0.52 GB (qwen3:0.6b); Q8_0 0.83 GB (qwen3:0.6b-q8_0).
Yes on 62 machines in this list. The smallest is M1 8GB: Q8_0, about 1.5 GB, roughly 66 tokens per second, via Ollama. ollama run qwen3:0.6b-q8_0
ollama run qwen3:0.6b-q8_0ollama run qwen3:0.6b-q8_0ollama run qwen3:0.6b-q8_0ollama run qwen3:0.6b-q8_0lms get unsloth/Qwen3-0.6B-GGUF@Q8_0llama-server -hf unsloth/Qwen3-0.6B-GGUF:Q8_0 -c 8192vllm serve Qwen/Qwen3-0.6B --max-model-len 8192 --gpu-memory-utilization 0.9| Hardware | Runtime | Quant | Memory | Fit | Speed | Command |
|---|---|---|---|---|---|---|
| M1 8GB | Ollama | Q8_0 | 1.5 GB | fits in fast memory | ~66 tok/s | ollama run qwen3:0.6b-q8_0 |
| M2 8GB | Ollama | Q8_0 | 1.5 GB | fits in fast memory | ~96 tok/s | ollama run qwen3:0.6b-q8_0 |
| M3 8GB | Ollama | Q8_0 | 1.5 GB | fits in fast memory | ~96 tok/s | ollama run qwen3:0.6b-q8_0 |
| RTX 3060 Ti 8GB | Ollama | Q8_0 | 1.5 GB | fits in fast memory | ~324 tok/s | ollama run qwen3:0.6b-q8_0 |
| RTX 3070 8GB | Ollama | Q8_0 | 1.5 GB | fits in fast memory | ~324 tok/s | ollama run qwen3:0.6b-q8_0 |
| RTX 4060 8GB | Ollama | Q8_0 | 1.5 GB | fits in fast memory | ~197 tok/s | ollama run qwen3:0.6b-q8_0 |
| RX 7600 8GB | Ollama | Q8_0 | 1.5 GB | fits in fast memory | ~208 tok/s | ollama run qwen3:0.6b-q8_0 |
| RTX 3080 10GB | Ollama | Q8_0 | 1.5 GB | fits in fast memory | ~549 tok/s | ollama run qwen3:0.6b-q8_0 |
| M1 16GB | Ollama | Q8_0 | 1.5 GB | fits in fast memory | ~66 tok/s | ollama run qwen3:0.6b-q8_0 |
| M2 16GB | Ollama | Q8_0 | 1.5 GB | fits in fast memory | ~96 tok/s | ollama run qwen3:0.6b-q8_0 |
| M3 16GB | Ollama | Q8_0 | 1.5 GB | fits in fast memory | ~96 tok/s | ollama run qwen3:0.6b-q8_0 |
| M4 16GB | Ollama | Q8_0 | 1.5 GB | fits in fast memory | ~116 tok/s | ollama run qwen3:0.6b-q8_0 |
| RTX 3060 12GB | Ollama | Q8_0 | 1.5 GB | fits in fast memory | ~260 tok/s | ollama run qwen3:0.6b-q8_0 |
| RTX 4070 12GB | Ollama | Q8_0 | 1.5 GB | fits in fast memory | ~364 tok/s | ollama run qwen3:0.6b-q8_0 |
| RTX 5070 12GB | Ollama | Q8_0 | 1.5 GB | fits in fast memory | ~486 tok/s | ollama run qwen3:0.6b-q8_0 |
| RTX 4060 Ti 16GB | Ollama | Q8_0 | 1.5 GB | fits in fast memory | ~208 tok/s | ollama run qwen3:0.6b-q8_0 |
| RTX 4070 Ti Super 16GB | Ollama | Q8_0 | 1.5 GB | fits in fast memory | ~486 tok/s | ollama run qwen3:0.6b-q8_0 |
| RTX 4080 16GB | Ollama | Q8_0 | 1.5 GB | fits in fast memory | ~518 tok/s | ollama run qwen3:0.6b-q8_0 |
| RTX 5080 16GB | Ollama | Q8_0 | 1.5 GB | fits in fast memory | ~694 tok/s | ollama run qwen3:0.6b-q8_0 |
| RX 7800 XT 16GB | Ollama | Q8_0 | 1.5 GB | fits in fast memory | ~451 tok/s | ollama run qwen3:0.6b-q8_0 |
| RX 9070 XT 16GB | Ollama | Q8_0 | 1.5 GB | fits in fast memory | ~463 tok/s | ollama run qwen3:0.6b-q8_0 |
| M2 24GB | Ollama | Q8_0 | 1.5 GB | fits in fast memory | ~96 tok/s | ollama run qwen3:0.6b-q8_0 |
| M3 24GB | Ollama | Q8_0 | 1.5 GB | fits in fast memory | ~96 tok/s | ollama run qwen3:0.6b-q8_0 |
| M4 24GB | Ollama | Q8_0 | 1.5 GB | fits in fast memory | ~116 tok/s | ollama run qwen3:0.6b-q8_0 |
| RTX 3090 24GB | Ollama | Q8_0 | 1.5 GB | fits in fast memory | ~677 tok/s | ollama run qwen3:0.6b-q8_0 |
| RTX 4090 24GB | Ollama | Q8_0 | 1.5 GB | fits in fast memory | ~729 tok/s | ollama run qwen3:0.6b-q8_0 |
| RX 7900 XTX 24GB | Ollama | Q8_0 | 1.5 GB | fits in fast memory | ~694 tok/s | ollama run qwen3:0.6b-q8_0 |
| M1 Pro 32GB | Ollama | Q8_0 | 1.5 GB | fits in fast memory | ~193 tok/s | ollama run qwen3:0.6b-q8_0 |
| M2 Pro 32GB | Ollama | Q8_0 | 1.5 GB | fits in fast memory | ~193 tok/s | ollama run qwen3:0.6b-q8_0 |
| M4 32GB | Ollama | Q8_0 | 1.5 GB | fits in fast memory | ~116 tok/s | ollama run qwen3:0.6b-q8_0 |
| M3 Pro 36GB | Ollama | Q8_0 | 1.5 GB | fits in fast memory | ~145 tok/s | ollama run qwen3:0.6b-q8_0 |
| Mac Studio M4 Max (32c GPU) 36GB | Ollama | Q8_0 | 1.5 GB | fits in fast memory | ~395 tok/s | ollama run qwen3:0.6b-q8_0 |
| Mac Studio M5 Max (32c GPU) 36GB | Ollama | Q8_0 | 1.5 GB | fits in fast memory | ~443 tok/s | ollama run qwen3:0.6b-q8_0 |
| RTX 5090 32GB | Ollama | Q8_0 | 1.5 GB | fits in fast memory | ~1295 tok/s | ollama run qwen3:0.6b-q8_0 |
| M4 Pro 48GB | Ollama | Q8_0 | 1.5 GB | fits in fast memory | ~263 tok/s | ollama run qwen3:0.6b-q8_0 |
| Mac Studio M4 Max (40c GPU) 48GB | Ollama | Q8_0 | 1.5 GB | fits in fast memory | ~526 tok/s | ollama run qwen3:0.6b-q8_0 |
| Mac Studio M5 Max (40c GPU) 48GB | Ollama | Q8_0 | 1.5 GB | fits in fast memory | ~592 tok/s | ollama run qwen3:0.6b-q8_0 |
| M1 Max 64GB | Ollama | Q8_0 | 1.5 GB | fits in fast memory | ~386 tok/s | ollama run qwen3:0.6b-q8_0 |
| M2 Max 64GB | Ollama | Q8_0 | 1.5 GB | fits in fast memory | ~386 tok/s | ollama run qwen3:0.6b-q8_0 |
| M3 Max 64GB | Ollama | Q8_0 | 1.5 GB | fits in fast memory | ~386 tok/s | ollama run qwen3:0.6b-q8_0 |
| M4 Max 64GB | Ollama | Q8_0 | 1.5 GB | fits in fast memory | ~526 tok/s | ollama run qwen3:0.6b-q8_0 |
| Mac Studio M4 Max (40c GPU) 64GB | Ollama | Q8_0 | 1.5 GB | fits in fast memory | ~526 tok/s | ollama run qwen3:0.6b-q8_0 |
| Mac Studio M5 Max (40c GPU) 64GB | Ollama | Q8_0 | 1.5 GB | fits in fast memory | ~592 tok/s | ollama run qwen3:0.6b-q8_0 |
| DGX Spark 64GB (ASUS, Dell, HP, MSI, Acer, Gigabyte) | Ollama | Q8_0 | 1.5 GB | fits in fast memory | ~247 tok/s | ollama run qwen3:0.6b-q8_0 |
| Mac Studio M3 Ultra 96GB | Ollama | Q8_0 | 1.5 GB | fits in fast memory | ~789 tok/s | ollama run qwen3:0.6b-q8_0 |
| Mac Studio M5 Ultra 96GB | Ollama | Q8_0 | 1.5 GB | fits in fast memory | ~1157 tok/s | ollama run qwen3:0.6b-q8_0 |
| M3 Max 128GB | Ollama | Q8_0 | 1.5 GB | fits in fast memory | ~386 tok/s | ollama run qwen3:0.6b-q8_0 |
| M4 Max 128GB | Ollama | Q8_0 | 1.5 GB | fits in fast memory | ~526 tok/s | ollama run qwen3:0.6b-q8_0 |
| Mac Studio M4 Max (40c GPU) 128GB | Ollama | Q8_0 | 1.5 GB | fits in fast memory | ~526 tok/s | ollama run qwen3:0.6b-q8_0 |
| Mac Studio M5 Max (40c GPU) 128GB | Ollama | Q8_0 | 1.5 GB | fits in fast memory | ~592 tok/s | ollama run qwen3:0.6b-q8_0 |
| DGX Spark 128GB (NVIDIA, ASUS, Dell, Lenovo, HP, MSI, Acer, Gigabyte) | Ollama | Q8_0 | 1.5 GB | fits in fast memory | ~247 tok/s | ollama run qwen3:0.6b-q8_0 |
| M2 Ultra 192GB | Ollama | Q8_0 | 1.5 GB | fits in fast memory | ~771 tok/s | ollama run qwen3:0.6b-q8_0 |
| 2x Mac Studio M5 Max (40c GPU) 128GB | llama.cpp | Q8_0 GGUF | 1.3 GB | fits in fast memory | ~652 tok/s | llama-server -hf unsloth/Qwen3-0.6B-GGUF:Q8_0 --rpc node2:50052 -c 8192 # run rpc-server on each other node |
| Mac Studio M3 Ultra 256GB | Ollama | Q8_0 | 1.5 GB | fits in fast memory | ~789 tok/s | ollama run qwen3:0.6b-q8_0 |
| Mac Studio M5 Ultra (64c GPU) 256GB | Ollama | Q8_0 | 1.5 GB | fits in fast memory | ~1157 tok/s | ollama run qwen3:0.6b-q8_0 |
| Mac Studio M5 Ultra (80c GPU) 256GB | Ollama | Q8_0 | 1.5 GB | fits in fast memory | ~1157 tok/s | ollama run qwen3:0.6b-q8_0 |
| 2x DGX Spark 128GB | llama.cpp | Q8_0 GGUF | 1.3 GB | fits in fast memory | ~272 tok/s | llama-server -hf unsloth/Qwen3-0.6B-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 | 1.3 GB | fits in fast memory | ~576 tok/s | llama-server -hf unsloth/Qwen3-0.6B-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 | 1.3 GB | fits in fast memory | ~1275 tok/s | llama-server -hf unsloth/Qwen3-0.6B-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 | 1.5 GB | fits in fast memory | ~789 tok/s | ollama run qwen3:0.6b-q8_0 |
| 4x DGX Spark 128GB | llama.cpp | Q8_0 GGUF | 1.3 GB | fits in fast memory | ~240 tok/s | llama-server -hf unsloth/Qwen3-0.6B-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 | 1.3 GB | fits in fast memory | ~1125 tok/s | llama-server -hf unsloth/Qwen3-0.6B-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