Home › GPUs & Macs › Apple M3 Pro (14-core GPU) vs Apple M3 Pro (18-core GPU)
Apple M3 Pro (14-core GPU) vs Apple M3 Pro (18-core GPU) for local LLMs
Across 5 tests measured on both, the two are about even (median ratio 0.99×).
On Llama 2 7B (Q4_0, llama-bench) the Apple M3 Pro (14-core GPU) generates 30.6 tokens/s and the Apple M3 Pro (18-core GPU) 30.7 tokens/s (measured).
Same model, same quant, same test
| Model | Quant / test | M3 Pro (14-core GPU) tok/s | M3 Pro (18-core GPU) tok/s | Faster | Sources and run details |
|---|---|---|---|---|---|
| Llama 2 7B | Q4_0 llama-bench | 30.6 Metal | 30.7 Metal | even | llama.cpp #4167 M3 Pro (14-core GPU) · llama.cpp #4167 M3 Pro (18-core GPU) |
| Llama 2 7B | Q8_0 llama-bench | 17.4 Metal | 17.5 Metal | even | llama.cpp #4167 M3 Pro (14-core GPU) · llama.cpp #4167 M3 Pro (18-core GPU) |
| Llama 3.1 8B Instruct | Q4_K_M LocalScore | 21.3 llamafile | 21.4 llamafile | even | LocalScore M3 Pro (14-core GPU) · LocalScore M3 Pro (18-core GPU) |
| Llama 3.2 1B Instruct | Q4_K_M LocalScore | 87.9 llamafile | 89.4 llamafile | even | LocalScore M3 Pro (14-core GPU) · LocalScore M3 Pro (18-core GPU) |
| Qwen2.5 14B Instruct | Q4_K_M LocalScore | 11.9 llamafile | 11.9 llamafile | even | LocalScore M3 Pro (14-core GPU) · LocalScore M3 Pro (18-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 M3 Pro (14-core GPU) | Apple M3 Pro (18-core GPU) | |
|---|---|---|
| Memory | Unified memory | Unified memory |
| Memory bandwidth | 150 GB/s | 150 GB/s |
All results for the Apple M3 Pro (14-core GPU) · All results for the Apple M3 Pro (18-core GPU) · More comparisons