Home › GPUs & Macs › Apple M1 Pro (14-core GPU) vs Apple M1 Pro (16-core GPU)
Apple M1 Pro (14-core GPU) vs Apple M1 Pro (16-core GPU) for local LLMs
Across 6 tests measured on both, the two are about even (median ratio 0.95×).
On Llama 2 7B (F16, llama-bench) the Apple M1 Pro (14-core GPU) generates 12.8 tokens/s and the Apple M1 Pro (16-core GPU) 12.8 tokens/s (measured).
Same model, same quant, same test
| Model | Quant / test | M1 Pro (14-core GPU) tok/s | M1 Pro (16-core GPU) tok/s | Faster | Sources and run details |
|---|---|---|---|---|---|
| Llama 2 7B | F16 llama-bench | 12.8 Metal | 12.8 Metal | even | llama.cpp #4167 M1 Pro (14-core GPU) · llama.cpp #4167 M1 Pro (16-core GPU) |
| Llama 2 7B | Q4_0 llama-bench | 35.5 Metal | 36.4 Metal | even | llama.cpp #4167 M1 Pro (14-core GPU) · llama.cpp #4167 M1 Pro (16-core GPU) |
| Llama 2 7B | Q8_0 llama-bench | 21.9 Metal | 22.3 Metal | even | llama.cpp #4167 M1 Pro (14-core GPU) · llama.cpp #4167 M1 Pro (16-core GPU) |
| Llama 3.1 8B Instruct | Q4_K_M LocalScore | 20.1 llamafile | 21.8 llamafile | M1 Pro (16-core GPU) +9% | LocalScore M1 Pro (14-core GPU) · LocalScore M1 Pro (16-core GPU) |
| Llama 3.2 1B Instruct | Q4_K_M LocalScore | 72.1 llamafile | 77.8 llamafile | M1 Pro (16-core GPU) +8% | LocalScore M1 Pro (14-core GPU) · LocalScore M1 Pro (16-core GPU) |
| Qwen2.5 14B Instruct | Q4_K_M LocalScore | 10.4 llamafile | 11.8 llamafile | M1 Pro (16-core GPU) +13% | LocalScore M1 Pro (14-core GPU) · LocalScore M1 Pro (16-core GPU) |
Generation speed in tokens/s, best measured backend for each chip (shown under the number). Backends can differ between the two chips; check them before drawing conclusions. Measured results only, nothing estimated.
Specs
| Apple M1 Pro (14-core GPU) | Apple M1 Pro (16-core GPU) | |
|---|---|---|
| Memory | Unified memory | Unified memory |
| Memory bandwidth | 200 GB/s | 200 GB/s |
All results for the Apple M1 Pro (14-core GPU) · All results for the Apple M1 Pro (16-core GPU) · More comparisons