MacBook Pro 14 M3 Pro
M3 Pro · 150 GB/s
Mac · Apple Silicon
M3 Pro with 18 GB: which local LLMs run?
With 18 GB of unified memory, the M3 Pro gives the GPU about 12 GB by default (150 GB/s): enough for models up to about Qwen3 14B.
- Unified memory18 GB
- GPU usable~12 GB
- Bandwidth150 GB/s
- GPU cores18
Unified memory is not VRAM: macOS reserves part of it. The limit can be raised with sudo sysctl iogpu.wired_limit_mb=…, at the cost of memory for apps. There is no offload to system RAM — it is the same memory.
- CPU cores
- 12
- GPU cores
- 18
- Bandwidth
- 150 GB/s
- Memory configurations
- 18 GB · 36 GB
- Sold in
- MacBook Pro 14, MacBook Pro 16
- Released
- October 2023
Official specData updated 24 days agoSource: Apple tech specs
What the M3 Pro with 18 GB runs
| Llama 3.1 8B InstructQ8_0 · 8K~10 GB · 11–16 tok/s estimate | Excellent |
|---|---|
| DeepSeek R1 Distill Llama 8BQ8_0 · 8K~10 GB · 11–16 tok/s estimate | Excellent |
| Qwen2.5 7B InstructQ8_0 · 8K~9 GB · 12–17 tok/s estimate | Excellent |
| Mistral 7B Instruct v0.3Q8_0 · 8K~9 GB · 12–17 tok/s estimate | Excellent |
| Gemma 3 4B ITQ8_0 · 8K~5.2 GB · 22–32 tok/s estimate | Excellent |
| Qwen3 4BQ8_0 · 8K~6 GB · 19–28 tok/s estimate | Excellent |
| Phi-4 Mini Instruct 3.8BQ8_0 · 8K~5.8 GB · 20–30 tok/s estimate | Excellent |
| Llama 3.2 3B InstructQ8_0 · 8K~4.9 GB · 24–36 tok/s estimate | Excellent |
| Qwen3 1.7BQ8_0 · 8K~3.5 GB · 39–60 tok/s estimate | Excellent |
| Llama 3.2 1B InstructQ8_0 · 8K~2.3 GB · 62–95 tok/s estimate | Excellent |
| Qwen3 0.6BQ8_0 · 8K~2.3 GB · 72–125 tok/s estimate | Excellent |
| Qwen3 14BQ4_K_M · 8K~10.8 GB · 10–15 tok/s estimate | Compatible |
| DeepSeek R1 Distill Qwen 14BQ4_K_M · 8K~11 GB · 9.5–14 tok/s estimate | Compatible |
| Qwen2.5 14B InstructQ4_K_M · 8K~11 GB · 9.5–14 tok/s estimate | Compatible |
| Phi-4 14BQ4_K_M · 8K~11 GB · 9.5–14 tok/s estimate | Compatible |
| Mistral Nemo 12B InstructQ5_K_M · 8K~10.5 GB · 10–15 tok/s estimate | Compatible |
| Gemma 3 12B ITQ6_K · 8K~11.1 GB · 9.5–14 tok/s estimate | Compatible |
| Gemma 2 9B ITQ6_K · 8K~10.3 GB · 11–16 tok/s estimate | Compatible |
| Qwen3 8BQ8_0 · 8K~10.3 GB · 11–15 tok/s estimate | Compatible |
| DeepSeek R1 (671B)Q8_0 · 8K~699 GB · No data | Incompatible |
| Qwen3 235B-A22BQ8_0 · 8K~247 GB · No data | Incompatible |
| gpt-oss-120bMXFP4 · 8K~63 GB · No data | Incompatible |
| Llama 4 Scout 17B-16E InstructQ8_0 · 8K~113 GB · No data | Incompatible |
| Qwen2.5 72B InstructQ8_0 · 8K~79.1 GB · No data | Incompatible |
| Llama 3.3 70B InstructQ8_0 · 8K~76.6 GB · No data | Incompatible |
| DeepSeek R1 Distill Llama 70BQ8_0 · 8K~76.6 GB · No data | Incompatible |
| Mixtral 8x7B InstructQ8_0 · 8K~50 GB · No data | Incompatible |
| Qwen3 32BQ8_0 · 8K~36.8 GB · No data | Incompatible |
| DeepSeek R1 Distill Qwen 32BQ8_0 · 8K~36.8 GB · No data | Incompatible |
| Qwen2.5 32B InstructQ8_0 · 8K~36.8 GB · No data | Incompatible |
| Qwen2.5 Coder 32B InstructQ8_0 · 8K~36.8 GB · No data | Incompatible |
| QwQ 32BQ8_0 · 8K~36.8 GB · No data | Incompatible |
| Qwen3 30B-A3BQ8_0 · 8K~33.2 GB · No data | Incompatible |
| Qwen3 Coder 30B-A3B InstructQ8_0 · 8K~33.2 GB · No data | Incompatible |
| Gemma 3 27B ITQ8_0 · 8K~30 GB · No data | Incompatible |
| Gemma 2 27B ITQ8_0 · 8K~31.4 GB · No data | Incompatible |
| Mistral Small 3.2 24B InstructQ8_0 · 8K~26.4 GB · No data | Incompatible |
| gpt-oss-20bMXFP4 · 8K~12.8 GB · No data | Incompatible |
For each model, the highest-precision quantization that runs entirely in fast memory; otherwise the one with the best result. Speeds are ranges; "estimate" unless a benchmark exists for this exact hardware.
EstimateData updated 24 days agoSource: Run LLM Locally engine
Where to buy
Prices and links follow the market you choose.
No current offer imported for this market: the link opens a marketplace search.