Which AI models can the Apple M2 Pro (32GB) run?
- 32 GB unified memory
- 200 GB/s bandwidth
- Apple
Every model we track, checked against the Apple M2 Pro (32GB) at 8K context.
| Model | Runs? | Best fit |
|---|---|---|
| Llama 3.2 1B | Yes | FP16 · ~3.3 GB · ~48 tok/s |
| Llama 3.2 3B | Yes | FP16 · ~8 GB · ~19 tok/s |
| Llama 3.1 8B | Yes | FP16 · ~17.7 GB · ~7 tok/s |
| Llama 3.1 70B | No | Needs ~46.1 GB at Q4_K_M; 21.4 GB available |
| Qwen2.5 7B | Yes | FP16 · ~16.3 GB · ~8 tok/s |
| Qwen2.5 14B | Yes | Q8_0 · ~17.8 GB · ~8 tok/s |
| Qwen2.5 32B | Yes, with heavy quantization | Q3_K_M · ~18.8 GB · ~7 tok/s |
| Qwen2.5 72B | No | Needs ~47.4 GB at Q4_K_M; 21.4 GB available |
| Qwen3 8B | Yes | FP16 · ~18.2 GB · ~7 tok/s |
| Qwen3 30B-A3B | Yes | Q4_K_M · ~19.9 GB · ~60 tok/s |
| DeepSeek R1 Distill Qwen 32B | Yes, with heavy quantization | Q3_K_M · ~18.8 GB · ~7 tok/s |
| Mistral 7B v0.3 | Yes | FP16 · ~16.2 GB · ~8 tok/s |
| Mistral Nemo 12B | Yes | Q8_0 · ~14.9 GB · ~9 tok/s |
| Mistral Small 24B | Yes | Q6_K · ~21.3 GB · ~6 tok/s |
| Mixtral 8x7B | Yes, with heavy quantization | Q2_K · ~19.2 GB · ~25 tok/s |
| Gemma 2 9B | Yes | Q8_0 · ~13.2 GB · ~12 tok/s |
| Gemma 2 27B | Yes | Q4_K_M · ~20.2 GB · ~7 tok/s |
| Gemma 3 4B | Yes | FP16 · ~10.3 GB · ~14 tok/s |
| Gemma 3 12B | Yes | Q8_0 · ~16.8 GB · ~9 tok/s |
| Gemma 3 27B | Yes | Q4_K_M · ~21.4 GB · ~7 tok/s |
| Phi-3.5 Mini 3.8B | Yes | FP16 · ~11.5 GB · ~16 tok/s |
| Phi-4 14B | Yes | Q8_0 · ~17.9 GB · ~8 tok/s |
| Qwen3 4B | Yes | FP16 · ~9.8 GB · ~15 tok/s |
| Qwen3 14B | Yes | Q8_0 · ~17.7 GB · ~8 tok/s |
| Qwen3 32B | Yes, with heavy quantization | Q3_K_M · ~18.8 GB · ~7 tok/s |
| Qwen3 Coder 30B-A3B | Yes | Q4_K_M · ~19.9 GB · ~60 tok/s |
| gpt-oss 20B | Yes | Q6_K · ~18.1 GB · ~41 tok/s |
| gpt-oss 120B | No | Needs ~72 GB at Q4_K_M; 21.4 GB available |
| Llama 4 Scout 17B-16E | No | Needs ~68.3 GB at Q4_K_M; 21.4 GB available |
Run it in the cloud
Need more than 32 GB? Rent a bigger GPU by the hour.
Rent a cloud GPU by the hour. No upfront cost, cancel any time.