Home › GPUs & Macs › Apple M3 Max (30-core GPU) vs Apple M5 Pro (16-core GPU)
Apple M3 Max (30-core GPU) vs Apple M5 Pro (16-core GPU) for local LLMs
Across 5 tests measured on both, the Apple M5 Pro (16-core GPU) is faster in 5 and generates a median 1.14× the tokens/s of the Apple M3 Max (30-core GPU).
On Llama 2 7B (F16, llama-bench) the Apple M3 Max (30-core GPU) generates 19.5 tokens/s and the Apple M5 Pro (16-core GPU) 21.4 tokens/s (measured).
Same model, same quant, same test
| Model | Quant / test | M3 Max (30-core GPU) tok/s | M5 Pro (16-core GPU) tok/s | Faster | Sources and run details |
|---|---|---|---|---|---|
| Llama 2 7B | F16 llama-bench | 19.5 Metal | 21.4 Metal | M5 Pro (16-core GPU) +9% | llama.cpp #4167 M3 Max (30-core GPU) · llama.cpp #4167 M5 Pro (16-core GPU) |
| Llama 2 7B | Q4_0 llama-bench | 56.6 Metal | 67.6 Metal | M5 Pro (16-core GPU) +19% | llama.cpp #4167 M3 Max (30-core GPU) · llama.cpp #4167 M5 Pro (16-core GPU) |
| Llama 2 7B | Q8_0 llama-bench | 34.3 Metal | 38.9 Metal | M5 Pro (16-core GPU) +14% | llama.cpp #4167 M3 Max (30-core GPU) · llama.cpp #4167 M5 Pro (16-core GPU) |
| Llama 3.1 8B Instruct | Q4_K_M LocalScore | 37.6 llamafile | 43.5 llamafile | M5 Pro (16-core GPU) +16% | LocalScore M3 Max (30-core GPU) · LocalScore M5 Pro (16-core GPU) |
| Qwen2.5 14B Instruct | Q4_K_M LocalScore | 20.1 llamafile | 22.1 llamafile | M5 Pro (16-core GPU) +10% | LocalScore M3 Max (30-core GPU) · LocalScore M5 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 M3 Max (30-core GPU) | Apple M5 Pro (16-core GPU) | |
|---|---|---|
| Memory | Unified memory | Unified memory |
| Memory bandwidth | 300 GB/s | 307 GB/s |
All results for the Apple M3 Max (30-core GPU) · All results for the Apple M5 Pro (16-core GPU) · More comparisons