Apple · M2 Pro · Mini PC
Best AI Models for Mac Mini M2 Pro (32 GB)
Runs 9B and 35B MoE models well; 14B and 27–31B fit but are slow (15 and 8 tok/s), and 70B doesn't fit in 28.5 GB.
| # | Model | Speed | Score |
|---|---|---|---|
| 1 | Qwen3.6 35B A3BAlibabaHow to run → Q4_K_M · 22.0 GB | 34.8 tok/s | |
| 2 | Gemma 4 26B A4B ITGoogle Q4_K_M · 16.5 GB | 39.0 tok/s | |
| 3 | Qwen3 30B A3B Thinking 2507Alibaba Q4_K_M · 19.4 GB | 41.2 tok/s | |
| 4 | GPT OSS 20BOpenAI Q4_K_M · 13.1 GB | 42.9 tok/s | |
| 5 | Q4_K_M · 19.3 GB | 48.2 tok/s |
| # | Model | Speed | Score |
|---|---|---|---|
| 1 | Qwen3.6 35B A3BAlibabaHow to run → Q4_K_M · 22.0 GB | 34.8 tok/s | |
| 2 | Gemma 4 26B A4B ITGoogle Q4_K_M · 16.5 GB | 39.0 tok/s | |
| 3 | GPT OSS 20BOpenAI Q4_K_M · 13.1 GB | 42.9 tok/s | |
| 4 | Q4_K_M · 7.9 GB | 26.4 tok/s | |
| 5 | Q4_K_M · 19.3 GB | 48.2 tok/s |
| # | Model | Speed | Score |
|---|---|---|---|
| 1 | Qwen3.6 35B A3BAlibabaHow to run → Q4_K_M · 22.0 GB | 34.8 tok/s | |
| 2 | Gemma 4 26B A4B ITGoogle Q4_K_M · 16.5 GB | 39.0 tok/s | |
| 3 | Qwen3.5 9BAlibaba Q4_K_M · 6.6 GB | 22.0 tok/s | |
| 4 | GPT OSS 20BOpenAI Q4_K_M · 13.1 GB | 42.9 tok/s | |
| 5 | Q4_K_M · 19.3 GB | 48.2 tok/s |
| # | Model | Speed | Score |
|---|---|---|---|
| 1 | Q4_K_M · 16.5 GB | 39.0 tok/s | |
| 2 | Qwen3 30B A3B Thinking 2507Alibaba Q4_K_M · 19.4 GB | 41.2 tok/s | |
| 3 | Qwen3.5 9BAlibaba Q4_K_M · 6.6 GB | 22.0 tok/s | |
| 4 | GPT OSS 20BOpenAI Q4_K_M · 13.1 GB | 42.9 tok/s | |
| 5 | Q4_K_M · 19.3 GB | 48.2 tok/s |
| # | Model | Speed ↓ | Score |
|---|---|---|---|
| 1 | GPT OSS 20BOpenAIHow to run → Q4_K_M · 13.1 GB | 42.9 tok/s | |
| 2 | Q4_K_M · 19.3 GB | 48.2 tok/s | |
| 3 | Qwen3 30B A3B Thinking 2507Alibaba Q4_K_M · 19.4 GB | 41.2 tok/s | |
| 4 | Qwen3 30B A3B Instruct 2507Alibaba Q4_K_M · 19.4 GB | 41.2 tok/s | |
| 5 | Gemma 4 26B A4B ITGoogle Q4_K_M · 16.5 GB | 39.0 tok/s |
The Mac Mini M2 Pro is a mid-range 32 GB mini PC: it runs 9B and 35B MoE models well. Among 32 GB machines, only the M1 Max, M2 Max and Orin write faster. macOS keeps about 3.5 GB, leaving 28.5 GB for models. Prompt reading is its weak spot: a 4K-token prompt waits 14 s before the first word on Qwen3.6 35B A3B. M2 Pro generation, released 2023; the newer M3 Pro writes -25% faster with 1.0× the compute.
Mac Mini M2 Pro strengths and weaknesses
Strengths
- Runs 35B MoE models: Qwen3.6 35B A3B at 34.8 tok/s in 22 GB
Weaknesses
- 70B models don't fit: Llama 3.3 70B Instruct needs 51.3 GB at Q4 vs 28.5 GB usable here
- Slow on large dense models: Qwen3.8 27B at 8.0 tok/s
- Slow prompt reading: A 32K-token prompt on Qwen3.6 35B A3B waits 1.9 min for the first token
- No CUDA: vLLM, AWQ, GPTQ and FP8 builds don't run; use GGUF or MLX
Memory for models
28.5 GB of 32 GB70B models need 51.3 GB at Q4, more than the 28.5 GB available.
Other memory sizes of the same machine first, then machines and GPUs people weigh against it, each with the one reason it is worth a look. Speeds on Qwen3.5 9B.
Output speed depends on memory bandwidth (200 GB/s). Grey rows need more than the 28.5 GB available.
- 9BQwen3.5 9B · Q4_K_M · 6.6 GB22.0tok/s
- 14BQwen3 14B · Q4_K_M · 11.9 GB14.7tok/s
- 8.0tok/s
- 35B MoEQwen3.6 35B A3B · Q4_K_M · 22.0 GB34.8tok/s
- 70BLlama 3.3 70B Instruct · does not fitneeds 51.3 GB
- 120B MoEGPT OSS 120B · does not fitneeds 70.8 GB
Estimated speeds: Q4_K_M, 4K context, one user. How we estimate
Pro generations: writing vs compute
M1 Pro → M5 ProWriting speed follows memory bandwidth and barely moved: 22.0 to 33.8 tok/s since the M1 Pro. Compute, which sets how fast prompts are read, grew 2.5×. Indexed to the M1 Pro; speeds on Qwen3.5 9B, same 32 GB.
All 1106 models on the Mac Mini M2 Pro
Featured onlyBest quant that fits in 28.5 GB, ordered by llmrun Score. Fit grades how well it runs here, S best. How ranking works
| Model | Quant | VRAM | Speed | Context | Score | Fit |
|---|---|---|---|---|---|---|
| BF16 | 18.5 GB | 8.8 t/s | — | B47, Decent | ||
Qwen3 14B BaseChat | Q4_K_M | 11.8 GB | 14.7 t/s | 33K | B54, Decent | |
Baichuan 7BChat | Q4_K_M | 5.0 GB | 30.3 t/s | 4K | A71, Runs well | |
Qwen2 1.5BChat | Q4_K_M | 1.7 GB | 109.4 t/s | 131K | S93, Runs great | |
| BF16 | 7.6 GB | 24.3 t/s | — | A65, Runs well | ||
| Q4_K_M | 11.0 GB | 17.3 t/s | 131K | B57, Decent | ||
Xgen 7B 8k BaseChat | Q4_K_M | 5.7 GB | 30.3 t/s | 8K | A71, Runs well | |
| Q4_K_M | 4.6 GB | 30.3 t/s | 2K | A71, Runs well | ||
OpenReasoning Nemotron 32BChatCodeReasoning | Q4_K_M | 24.3 GB | 6.8 t/s | 131K | C37, Usable | |
Qwen2.5 Coder 3BChatCodeFunctions | Q4_K_M | 2.8 GB | 62.8 t/s | 33K | A84, Runs well | |
Yi 34BChat | Q4_K_M | 21.9 GB | 6.5 t/s | 4K | C43, Usable | |
| IQ2_S | 27.6 GB | 28.2 t/s | 33K | C43, Usable | ||
DeepSeek Coder v2 Lite BaseChatCode | Q4_K_M | 10.2 GB | 56.4 t/s | 164K | A82, Runs well | |
| Q4_K_M | 10.5 GB | 57.9 t/s | 8K | A83, Runs well | ||
| Q4_K_M | 6.2 GB | 26.0 t/s | 8K | A66, Runs well |
Specs
- Chip
- M2 Pro
- CPU cores
- 12
- GPU cores
- 19
- Unified memory
- 32 GB
- For models
- 28.5 GB (macOS keeps 3.5 GB)
- Memory bandwidth
- 200 GB/s
- FP16 (dense)
- 13.6 TFLOPS
- Neural Engine
- 16 TOPS
- Released
- 2023-01-24
Mac Mini M2 Pro FAQ
How much memory is available for AI on the Mac Mini M2 Pro (32 GB)?
About 28.5 GB of the 32 GB. macOS keeps roughly 3.5 GB for itself and other apps.
What is the largest model the Mac Mini M2 Pro (32 GB) can run?
Qwen3.6 35B A3B (36.0B) at Q4_K_M uses 22.0 GB, 77% of the available memory, and runs at about 34.8 tok/s.
Can the Mac Mini M2 Pro (32 GB) run 70B models?
No. Llama 3.3 70B Instruct needs 51.3 GB at Q4, more than the 28.5 GB available.
How long does the Mac Mini M2 Pro (32 GB) take to read a long prompt?
On Qwen3.6 35B A3B, about 14 s for a 4K-token prompt and 1.9 min for a 32K-token document before the first word appears.
Is renting a GPU cheaper than buying the Mac Mini M2 Pro?
An RTX PRO 4500 Blackwell (32 GB) rents from $0.34/hr.