Apple
Best local LLMs for the Mac · M4 Pro, 48 GB
48 GB unified memory · 273 GB/s · ~35 GB usable. A great mid-range local-LLM Mac — runs 32B comfortably, 70B at a squeeze.
Try it with your context & use case
Preset to the Mac · M4 Pro, 48 GB. Change the context length or filter by use case.
24 models run well and 3 run tight on Mac · M4 Pro, 48 GB (any context — just fitting the weights) — 35 GB usable (≈72% of unified memory).
| Model | Size | Fit | Best quant | Needs | Memory | Speed |
|---|---|---|---|---|---|---|
| Mixtral 8x7B (MoE)MoE | 46.7B | Runs well | Q5_K_M | 33 GB | ~14 tok/s | |
| Qwen3 32B | 32.8B | Runs well | Q8_0 | 35 GB | ~5.6 tok/s | |
| Qwen2.5 32B | 32.8B | Runs well | Q8_0 | 35 GB | ~5.6 tok/s | |
| Qwen2.5-Coder 32B | 32.8B | Runs well | Q8_0 | 35 GB | ~5.6 tok/s | |
| DeepSeek-R1 Distill Qwen 32B | 32.8B | Runs well | Q8_0 | 35 GB | ~5.6 tok/s | |
| Qwen3 30B-A3B (MoE)MoE | 30.5B | Runs well | Q8_0 | 32 GB | ~36 tok/s | |
| Gemma 3 27B | 27.4B | Runs well | Q8_0 | 29 GB | ~6.8 tok/s | |
| Gemma 2 27B | 27.2B | Runs well | Q8_0 | 29 GB | ~6.8 tok/s | |
| Mistral Small 3 24B | 23.6B | Runs well | Q8_0 | 25 GB | ~7.8 tok/s | |
| Qwen3 14B | 14.8B | Runs well | FP16 | 29 GB | ~6.6 tok/s | |
| Qwen2.5 14B | 14.8B | Runs well | FP16 | 29 GB | ~6.6 tok/s | |
| DeepSeek-R1 Distill Qwen 14B | 14.8B | Runs well | FP16 | 29 GB | ~6.6 tok/s | |
| Phi-4 14B | 14.7B | Runs well | FP16 | 29 GB | ~6.7 tok/s | |
| Mistral Nemo 12B | 12.2B | Runs well | FP16 | 24 GB | ~8.1 tok/s | |
| Gemma 3 12B | 12.2B | Runs well | FP16 | 24 GB | ~8.1 tok/s | |
| Gemma 2 9B | 9.2B | Runs well | FP16 | 19 GB | ~11 tok/s | |
| Qwen3 8B | 8.2B | Runs well | FP16 | 17 GB | ~12 tok/s | |
| Llama 3.1 8B | 8.0B | Runs well | FP16 | 16 GB | ~12 tok/s | |
| Qwen2.5-Coder 7B | 7.6B | Runs well | FP16 | 15 GB | ~13 tok/s | |
| Mistral 7B v0.3 | 7.3B | Runs well | FP16 | 15 GB | ~14 tok/s | |
| Gemma 3 4B | 4.3B | Runs well | FP16 | 9.1 GB | ~23 tok/s | |
| Qwen3 4B | 4B | Runs well | FP16 | 8.5 GB | ~25 tok/s | |
| Llama 3.2 3B | 3.2B | Runs well | FP16 | 7.0 GB | ~31 tok/s | |
| Llama 3.2 1B | 1.2B | Runs well | FP16 | 3.2 GB | ~79 tok/s | |
| Qwen2.5 72B | 72.7B | Runs (tight) | Q2_K | 30 GB | ~6.5 tok/s | |
| Llama 3.3 70B | 70.6B | Runs (tight) | Q3_K_M | 34 GB | ~5.7 tok/s | |
| DeepSeek-R1 Distill Llama 70B | 70.6B | Runs (tight) | Q3_K_M | 34 GB | ~5.7 tok/s | |
| DeepSeek-R1 671B-A37B (MoE)MoE | 671B | Won't fit | — | 273 GB | — | |
| Qwen3 235B-A22B (MoE)MoE | 235B | Won't fit | — | 96 GB | — | |
| Llama 4 Scout (MoE)MoE | 109B | Won't fit | — | 45 GB | — |
Estimates, computed in your browser — VRAM, quantisation and speed vary with your runtime and settings. How we estimate →
Top picks for the Mac · M4 Pro, 48 GB
Every model on the Mac · M4 Pro, 48 GB
| Model | Size | Fit | Best quant | Needs | Speed |
|---|---|---|---|---|---|
| Mixtral 8x7B (MoE)MoE | 46.7B | Runs well | Q5 | 34 GB | ~14 tok/s |
| Qwen3 32B | 32.8B | Runs well | Q6 | 29 GB | ~7.3 tok/s |
| Qwen2.5 32B | 32.8B | Runs well | Q6 | 29 GB | ~7.3 tok/s |
| Qwen2.5-Coder 32B | 32.8B | Runs well | Q6 | 29 GB | ~7.3 tok/s |
| DeepSeek-R1 Distill Qwen 32B | 32.8B | Runs well | Q6 | 29 GB | ~7.3 tok/s |
| Qwen3 30B-A3B (MoE)MoE | 30.5B | Runs well | Q8 | 33 GB | ~36 tok/s |
| Gemma 3 27B | 27.4B | Runs well | Q8 | 33 GB | ~6.8 tok/s |
| Gemma 2 27B | 27.2B | Runs well | Q8 | 32 GB | ~6.8 tok/s |
| Mistral Small 3 24B | 23.6B | Runs well | Q8 | 26 GB | ~7.8 tok/s |
| Qwen3 14B | 14.8B | Runs well | FP16 | 31 GB | ~6.6 tok/s |
| Qwen2.5 14B | 14.8B | Runs well | FP16 | 31 GB | ~6.6 tok/s |
| DeepSeek-R1 Distill Qwen 14B | 14.8B | Runs well | FP16 | 31 GB | ~6.6 tok/s |
| Phi-4 14B | 14.7B | Runs well | FP16 | 31 GB | ~6.7 tok/s |
| Mistral Nemo 12B | 12.2B | Runs well | FP16 | 26 GB | ~8.1 tok/s |
| Gemma 3 12B | 12.2B | Runs well | FP16 | 27 GB | ~8.1 tok/s |
| Gemma 2 9B | 9.2B | Runs well | FP16 | 21 GB | ~11 tok/s |
| Qwen3 8B | 8.2B | Runs well | FP16 | 18 GB | ~12 tok/s |
| Llama 3.1 8B | 8.0B | Runs well | FP16 | 17 GB | ~12 tok/s |
| Qwen2.5-Coder 7B | 7.6B | Runs well | FP16 | 16 GB | ~13 tok/s |
| Mistral 7B v0.3 | 7.3B | Runs well | FP16 | 16 GB | ~14 tok/s |
| Gemma 3 4B | 4.3B | Runs well | FP16 | 10 GB | ~23 tok/s |
| Qwen3 4B | 4B | Runs well | FP16 | 9.6 GB | ~25 tok/s |
| Llama 3.2 3B | 3.2B | Runs well | FP16 | 7.8 GB | ~31 tok/s |
| Llama 3.2 1B | 1.2B | Runs well | FP16 | 3.4 GB | ~79 tok/s |
| Qwen2.5 72B | 72.7B | Runs (tight) | Q2 | 33 GB | ~6.5 tok/s |
| Llama 3.3 70B | 70.6B | Runs (tight) | Q2 | 32 GB | ~6.6 tok/s |
| DeepSeek-R1 Distill Llama 70B | 70.6B | Runs (tight) | Q2 | 32 GB | ~6.6 tok/s |
| DeepSeek-R1 671B-A37B (MoE)MoE | 671B | Won't fit | — | 303 GB | — |
| Qwen3 235B-A22B (MoE)MoE | 235B | Won't fit | — | 98 GB | — |
| Llama 4 Scout (MoE)MoE | 109B | Won't fit | — | 46 GB | — |
Similar hardware
FAQ
What is the best LLM for the Mac · M4 Pro, 48 GB?
For general use, Mixtral 8x7B (MoE) is the strongest model that runs well on the Mac · M4 Pro, 48 GB. See the picks-by-use-case below for coding, reasoning and more.
How much can the Mac · M4 Pro, 48 GB run?
The Mac · M4 Pro, 48 GB has 48 GB of unified memory, of which about 35 GB is usable for a model. That runs 24 of our tracked models well and 3 more at a tight quantisation.
Estimates at 8K context — see how we compute these.