Home › GPUs & Macs › Apple M1 Max (32-core GPU) vs Apple M2 Max (30-core GPU)
Apple M1 Max (32-core GPU) vs Apple M2 Max (30-core GPU) for local LLMs
Across 6 tests measured on both, the two are about even (median ratio 1.00×).
On Llama 2 7B (F16, llama-bench) the Apple M1 Max (32-core GPU) generates 23.0 tokens/s and the Apple M2 Max (30-core GPU) 24.2 tokens/s (measured).
Same model, same quant, same test
| Model | Quant / test | M1 Max (32-core GPU) tok/s | M2 Max (30-core GPU) tok/s | Faster | Sources and run details |
|---|---|---|---|---|---|
| Llama 2 7B | F16 llama-bench | 23.0 Metal | 24.2 Metal | M2 Max (30-core GPU) +5% | llama.cpp #4167 M1 Max (32-core GPU) · llama.cpp #4167 M2 Max (30-core GPU) |
| Llama 2 7B | Q4_0 llama-bench | 61.2 Metal | 61.0 Metal | even | llama.cpp #4167 M1 Max (32-core GPU) · llama.cpp #4167 M2 Max (30-core GPU) |
| Llama 2 7B | Q8_0 llama-bench | 40.2 Metal | 40.0 Metal | even | llama.cpp #4167 M1 Max (32-core GPU) · llama.cpp #4167 M2 Max (30-core GPU) |
| Llama 3.1 8B Instruct | Q4_K_M LocalScore | 36.2 llamafile | 36.2 llamafile | even | LocalScore M1 Max (32-core GPU) · LocalScore M2 Max (30-core GPU) |
| Llama 3.2 1B Instruct | Q4_K_M LocalScore | 121 llamafile | 128 llamafile | M2 Max (30-core GPU) +5% | LocalScore M1 Max (32-core GPU) · LocalScore M2 Max (30-core GPU) |
| Qwen2.5 14B Instruct | Q4_K_M LocalScore | 19.8 llamafile | 14.5 llamafile | M1 Max (32-core GPU) +37% | LocalScore M1 Max (32-core GPU) · LocalScore M2 Max (30-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 Max (32-core GPU) | Apple M2 Max (30-core GPU) | |
|---|---|---|
| Memory | Unified memory | Unified memory |
| Memory bandwidth | 400 GB/s | 400 GB/s |
All results for the Apple M1 Max (32-core GPU) · All results for the Apple M2 Max (30-core GPU) · More comparisons