Apple · M3 Pro · Laptop
Best AI Models for MacBook Pro 14" M3 Pro (36 GB)
Runs 9B and 35B MoE models well; 14B and 27–31B fit but are slow (11 and 6 tok/s), and 70B doesn't fit in 32.5 GB.
| # | Model | Speed | Score |
|---|---|---|---|
| 1 | Qwen3.6 35B A3BAlibabaHow to run → Q4_K_M · 22.0 GB | 28.1 tok/s | |
| 2 | Gemma 4 26B A4B ITGoogle Q4_K_M · 16.5 GB | 31.1 tok/s | |
| 3 | Qwen3 30B A3B Thinking 2507Alibaba Q4_K_M · 19.4 GB | 34.3 tok/s | |
| 4 | GPT OSS 20BOpenAI Q4_K_M · 13.1 GB | 33.9 tok/s | |
| 5 | Q4_K_M · 19.3 GB | 41.4 tok/s |
| # | Model | Speed | Score |
|---|---|---|---|
| 1 | Qwen3.6 35B A3BAlibabaHow to run → Q4_K_M · 22.0 GB | 28.1 tok/s | |
| 2 | Gemma 4 26B A4B ITGoogle Q4_K_M · 16.5 GB | 31.1 tok/s | |
| 3 | GPT OSS 20BOpenAI Q4_K_M · 13.1 GB | 33.9 tok/s | |
| 4 | Q4_K_M · 7.9 GB | 19.8 tok/s | |
| 5 | Q4_K_M · 19.3 GB | 41.4 tok/s |
| # | Model | Speed | Score |
|---|---|---|---|
| 1 | Qwen3.6 35B A3BAlibabaHow to run → Q4_K_M · 22.0 GB | 28.1 tok/s | |
| 2 | Gemma 4 26B A4B ITGoogle Q4_K_M · 16.5 GB | 31.1 tok/s | |
| 3 | Qwen3.5 9BAlibaba Q4_K_M · 6.6 GB | 16.5 tok/s | |
| 4 | GPT OSS 20BOpenAI Q4_K_M · 13.1 GB | 33.9 tok/s | |
| 5 | Q4_K_M · 19.3 GB | 41.4 tok/s |
| # | Model | Speed ↓ | Score |
|---|---|---|---|
| 1 | Q4_K_M · 19.4 GB | 34.3 tok/s | |
| 2 | Q4_K_M · 19.3 GB | 41.4 tok/s | |
| 3 | Qwen3 30B A3B Instruct 2507Alibaba Q4_K_M · 19.4 GB | 34.3 tok/s | |
| 4 | GPT OSS 20BOpenAI Q4_K_M · 13.1 GB | 33.9 tok/s | |
| 5 | Gemma 4 26B A4B ITGoogle Q4_K_M · 16.5 GB | 31.1 tok/s |
The MacBook Pro 14" M3 Pro is a 36 GB laptop with more memory than speed: it runs 9B and 35B MoE models well. Among 36 GB machines, only the M5 Max, M4 Max and M3 Max write faster. macOS keeps about 3.5 GB, leaving 32.5 GB for models. Prompt reading is its weak spot: a 4K-token prompt waits 15 s before the first word on Qwen3.6 35B A3B. M3 Pro generation, released 2023; the newer M4 Pro writes 82% faster with 1.3× the compute.
MacBook Pro 14" M3 Pro strengths and weaknesses
Strengths
- Runs 35B MoE models: Qwen3.6 35B A3B at 28.1 tok/s in 22 GB
Weaknesses
- 70B models don't fit: Llama 3.3 70B Instruct needs 51.3 GB at Q4 vs 32.5 GB usable here
- Slow on large dense models: Qwen3.8 27B at 6.0 tok/s
- Slow prompt reading: A 32K-token prompt on Qwen3.6 35B A3B waits 2.0 min for the first token
- No CUDA: vLLM, AWQ, GPTQ and FP8 builds don't run; use GGUF or MLX
Memory for models
32.5 GB of 36 GB70B models need 51.3 GB at Q4, more than the 32.5 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 (150 GB/s). Grey rows need more than the 32.5 GB available.
- 9BQwen3.5 9B · Q4_K_M · 6.6 GB16.5tok/s
- 14BQwen3 14B · Q4_K_M · 11.9 GB11.1tok/s
- 6.0tok/s
- 35B MoEQwen3.6 35B A3B · Q4_K_M · 22.0 GB28.1tok/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 1115 models on the MacBook Pro 14" M3 Pro
Featured onlyBest quant that fits in 32.5 GB, ordered by llmrun Score. Fit grades how well it runs here, S best. How ranking works
| Model | Quant | VRAM | Speed | Context | Score | Fit |
|---|---|---|---|---|---|---|
| Q4_K_M | 7.5 GB | 14.7 t/s | 4K | B54, Decent | ||
DeepSeek v2 LiteChat | Q4_K_M | 10.2 GB | 45.8 t/s | 164K | A78, Runs well | |
| Q4_K_M | 6.0 GB | 28.5 t/s | — | A69, Runs well | ||
Luciole 8B BaseChat | Q4_K_M | 5.6 GB | 18.8 t/s | 131K | B58, Decent | |
Qwen2.5 Coder 3B InstructChatCodeFunctions | Q4_K_M | 2.8 GB | 47.1 t/s | 33K | A78, Runs well | |
Neural Chat 7B v3 3ChatMath | Q4_K_M | 6.8 GB | 21.4 t/s | 33K | B62, Decent | |
| BF16 | 5.9 GB | 17.7 t/s | 2K | B58, Decent | ||
| Q4_K_M | 12.1 GB | 46.7 t/s | 33K | A78, Runs well | ||
| Q4_K_M | 14.2 GB | 45.3 t/s | 131K | A77, Runs well | ||
XiYanSQL QwenCoder 32B 2504ChatCode | Q4_K_M | 23.8 GB | 5.2 t/s | 33K | C42, Usable | |
| Q4_K_M | 11.8 GB | 11.1 t/s | 41K | B49, Decent | ||
| BF16 | 18.6 GB | 6.3 t/s | 33K | C43, Usable | ||
| Q4_K_M | 4.8 GB | 39.8 t/s | — | A75, Runs well | ||
Gemma 3 12B PtVision | Q4_K_M | 10.9 GB | 13.1 t/s | — | B52, Decent | |
| Q4_K_M | 7.3 GB | 19.5 t/s | 131K | B60, Decent |
Specs
- Chip
- M3 Pro
- CPU cores
- 12
- GPU cores
- 18
- Unified memory
- 36 GB
- For models
- 32.5 GB (macOS keeps 3.5 GB)
- Memory bandwidth
- 150 GB/s
- FP16 (dense)
- 12.9 TFLOPS
- Neural Engine
- 18 TOPS
- Released
- 2023-11-07
MacBook Pro 14" M3 Pro FAQ
How much memory is available for AI on the MacBook Pro 14" M3 Pro (36 GB)?
About 32.5 GB of the 36 GB. macOS keeps roughly 3.5 GB for itself and other apps.
What is the largest model the MacBook Pro 14" M3 Pro (36 GB) can run?
Qwen3.6 35B A3B (36.0B) at Q4_K_M uses 22.0 GB, 68% of the available memory, and runs at about 28.1 tok/s.
Can the MacBook Pro 14" M3 Pro (36 GB) run 70B models?
No. Llama 3.3 70B Instruct needs 51.3 GB at Q4, more than the 32.5 GB available.
How long does the MacBook Pro 14" M3 Pro (36 GB) take to read a long prompt?
On Qwen3.6 35B A3B, about 15 s for a 4K-token prompt and 2.0 min for a 32K-token document before the first word appears.
Is renting a GPU cheaper than buying the MacBook Pro 14" M3 Pro?
An A40 (48 GB) rents from $0.35/hr.