Apple · M3 Pro · Laptop
Best AI Models for MacBook Pro 14" M3 Pro (18 GB)
Runs 9B models well; dense 14B fits but is slow at 11 tok/s, and 27–31B only just fit in 14.5 GB.
| # | Model | Speed | Score |
|---|---|---|---|
| 1 | Qwen3.5 9BAlibabaHow to run → Q4_K_M · 6.6 GB | 16.5 tok/s | |
| 2 | GPT OSS 20BOpenAI Q4_K_M · 13.1 GB | 33.9 tok/s | |
| 3 | Qwen3 8BAlibaba Q4_K_M · 7.6 GB | 19.0 tok/s | |
| 4 | Gemma 3 4B ITGoogle Q4_K_M · 5.0 GB | 37.0 tok/s | |
| 5 | Gemma 4 E4B ITGoogle Q4_K_M · 5.6 GB | 19.7 tok/s |
| # | Model | Speed | Score |
|---|---|---|---|
| 1 | GPT OSS 20BOpenAIHow to run → Q4_K_M · 13.1 GB | 33.9 tok/s | |
| 2 | Q4_K_M · 7.9 GB | 19.8 tok/s | |
| 3 | Qwen3.5 9BAlibaba Q4_K_M · 6.6 GB | 16.5 tok/s | overall |
| 4 | Qwen3.5 4BAlibaba Q4_K_M · 3.4 GB | 32.2 tok/s | |
| 5 | Gemma 4 E4B ITGoogle Q4_K_M · 5.6 GB | 19.7 tok/s |
| # | Model | Speed | Score |
|---|---|---|---|
| 1 | Qwen3.5 9BAlibabaHow to run → Q4_K_M · 6.6 GB | 16.5 tok/s | |
| 2 | Q4_K_M · 7.9 GB | 19.8 tok/s | |
| 3 | Q4_K_M · 2.3 GB | 128 tok/s | |
| 4 | GPT OSS 20BOpenAI Q4_K_M · 13.1 GB | 33.9 tok/s | |
| 5 | Qwen3.5 4BAlibaba Q4_K_M · 3.4 GB | 32.2 tok/s |
| # | Model | Speed ↓ | Score |
|---|---|---|---|
| 1 | GPT OSS 20BOpenAIHow to run → Q4_K_M · 13.1 GB | 33.9 tok/s | |
| 2 | Ling 3.0 TinyInclusion AI Q4_K_M · 5.5 GB | 73.0 tok/s | |
| 3 | Gemma 4 E4B ITGoogle Q4_K_M · 5.6 GB | 19.7 tok/s | |
| 4 | Qwen3 8BAlibaba Q4_K_M · 7.6 GB | 19.0 tok/s | |
| 5 | Qwen3.5 9BAlibaba Q4_K_M · 6.6 GB | 16.5 tok/s |
| # | Model | Speed | Score |
|---|---|---|---|
| 1 | Qwen3.5 9BAlibabaHow to run → Q4_K_M · 6.6 GB | 16.5 tok/s | |
| 2 | Gemma 3 4B ITGoogle Q4_K_M · 5.0 GB | 37.0 tok/s | |
| 3 | Qwen3.5 4BAlibaba Q4_K_M · 3.4 GB | 32.2 tok/s | |
| 4 | Q4_K_M · 6.5 GB | 16.9 tok/s | |
| 5 | Cosmos Reason2 2BNVIDIA Q4_K_M · 3.5 GB | 65.2 tok/s |
The MacBook Pro 14" M3 Pro runs 9B models well. macOS keeps about 3.5 GB, leaving 14.5 GB for models. Prompt reading is its weak spot: a 4K-token prompt waits 39 s before the first word on Qwen3.5 9B. M3 Pro generation, released 2023; the newer M4 Pro writes 82% faster with 1.7× the compute.
MacBook Pro 14" M3 Pro strengths and weaknesses
Strengths
- Runs 7–9B models: Qwen3.5 9B at 16.5 tok/s in 6.6 GB
Weaknesses
- 35B MoE models don't fit: Qwen3.6 35B A3B needs 22 GB at Q4 vs 14.5 GB usable here
- Slow on large dense models: Qwen3.8 27B at 7.9 tok/s
- Slow prompt reading: A 32K-token prompt on Qwen3.5 9B waits 5.3 min for the first token
- No CUDA: vLLM, AWQ, GPTQ and FP8 builds don't run; use GGUF or MLX
Memory for models
14.5 GB of 18 GB35B MoE models need 22 GB at Q4, more than the 14.5 GB available. The 36 GB config fits Qwen3.8 27B →
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 14.5 GB available.
- 9BQwen3.5 9B · Q4_K_M · 6.6 GB16.5tok/s
- 14BQwen3 14B · Q4_K_M · 11.9 GB11.1tok/s
- 7.9tok/s
- 35B MoEQwen3.6 35B A3B · does not fitneeds 22.0 GB
- 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 2.8×. Indexed to the M1 Pro; speeds on Qwen3.5 9B, same 16 GB.
All 975 models on the MacBook Pro 14" M3 Pro
Featured onlyBest quant that fits in 14.5 GB, ordered by llmrun Score. Fit grades how well it runs here, S best. How ranking works
| Model | Quant | VRAM | Speed | Context | Score | Fit |
|---|---|---|---|---|---|---|
Qwen2 7BChat | Q4_K_M | 5.8 GB | 21.0 t/s | 131K | B61, Decent | |
| Q3_K_S | 14.5 GB | 41.7 t/s | 33K | C44, Usable | ||
Qwen 7B ChatChat | Q4_K_M | 7.6 GB | 20.6 t/s | 33K | B60, Decent | |
| Q4_K_M | 7.3 GB | 22.1 t/s | — | B62, Decent | ||
| Q4_K_M | 7.9 GB | 19.0 t/s | 131K | B59, Decent | ||
Qwen1.5 1.8BChat | Q4_K_M | 4.6 GB | 58.3 t/s | 33K | A83, Runs well | |
GhostFace 24B V1Chat | Q2_K | 13.7 GB | 9.8 t/s | 131K | D27, Barely runs | |
Gemma 2 9BChat | Q4_K_M | 8.8 GB | 17.2 t/s | — | B57, Decent | |
Gemma 2 27B ITChat | IQ3_S | 14.3 GB | 8.2 t/s | 8K | D17, Barely runs | |
Qwen2.5 Coder 7BChatCodeFunctions | Q4_K_M | 5.8 GB | 21.0 t/s | 33K | B61, Decent | |
GLM Z1 32B 0414ChatFunctions | Q2_K_S | 14.3 GB | 7.8 t/s | 33K | D16, Barely runs | |
| Q4_K_M | 7.9 GB | 19.0 t/s | 131K | B59, Decent | ||
Yi 1.5 34B ChatChat | IQ2_M | 12.9 GB | 8.5 t/s | 4K | C34, Usable | |
Qwen2.5 VL 32B InstructVision | IQ2_XXS | 13.8 GB | 10.5 t/s | 128K | D26, Barely runs | |
| Q2_K | 13.7 GB | 9.8 t/s | 33K | D27, Barely runs |
Specs
- Chip
- M3 Pro
- CPU cores
- 11
- GPU cores
- 14
- Unified memory
- 18 GB
- For models
- 14.5 GB (macOS keeps 3.5 GB)
- Memory bandwidth
- 150 GB/s
- FP16 (dense)
- 10.0 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 (18 GB)?
About 14.5 GB of the 18 GB. macOS keeps roughly 3.5 GB for itself and other apps.
What is the largest model the MacBook Pro 14" M3 Pro (18 GB) can run?
GPT OSS 20B (20.9B) at Q4_K_M uses 13.1 GB, 91% of the available memory, and runs at about 33.9 tok/s.
Can the MacBook Pro 14" M3 Pro (18 GB) run Qwen3.5 9B?
Yes. Qwen3.5 9B fits at Q4_K_M, using 6.6 GB of the 14.5 GB available, and it is the best model that runs well on this machine.
Can the MacBook Pro 14" M3 Pro (18 GB) run 35B MoE models?
No. Qwen3.6 35B A3B needs 22.0 GB at Q4, more than the 14.5 GB available.
How long does the MacBook Pro 14" M3 Pro (18 GB) take to read a long prompt?
On Qwen3.5 9B, about 39 s for a 4K-token prompt and 5.3 min for a 32K-token document before the first word appears.
Is renting a GPU cheaper than buying the MacBook Pro 14" M3 Pro?
A Tesla V100 (16 GB) rents from $0.10/hr.