2benchesM4 Max
A18 · M4 · 2024 CPU family · 3 nm · First available 2024-11-08
Every configuration
| Configuration | CPU core layout | Bandwidth | GB7 CPU | GB7 Metal | Prefill Q4_0 tok/s | Decode Q4_0 tok/s |
|---|---|---|---|---|---|---|
| 14 CPU / 32 GPU (binned) | 10P + 4E | 410 GB/s | 25,450 | 168,673 | 713.9 | 70 |
| 16 CPU / 40 GPU (full chip) | 12P + 4E | 546 GB/s | 28,755 | 203,884 | 885.7 | 83.1 |
All benchmarks
| Benchmark | Median |
|---|---|
| Geekbench 7 · Multi-core | 26,979 (25,290–29,076, 6 sources) |
| Geekbench 7 · Metal GPU | 186,399 (168,265–207,414, 6 sources) |
| Geekbench AI · GPU · Single | — |
| Geekbench AI · GPU · Half | — |
| Geekbench AI · GPU · Quantized | — |
| Geekbench AI · Neural Engine · Single | — |
| Geekbench AI · Neural Engine · Half | — |
| Geekbench AI · Neural Engine · Quantized | — |
| Prefill 512 · F16 | 829.5 (736.3–922.8, 2 sources) |
| Decode 128 · F16 | 28 (24.3–31.6, 2 sources) |
| Prefill 512 · Q8_0 | 805.3 (718.6–891.9, 2 sources) |
| Decode 128 · Q8_0 | 49 (43.9–54.1, 2 sources) |
| Prefill 512 · Q4_0 | 799.8 (713.9–885.7, 2 sources) |
| Decode 128 · Q4_0 | 76.5 (70–83.1, 2 sources) |
Similar performance
Nearest measured configurations by Geekbench 7 CPU and Metal. Ratios above 1 are faster.
14 CPU / 32 GPU (binned)
- M3 Max 16 CPU / 40 GPU (full chip), 2023-11: CPU 0.96×, GPU 0.97×
- M3 Max 14 CPU / 30 GPU (binned), 2023-11: CPU 0.87×, GPU 0.78×
- M2 Ultra 24 CPU / 76 GPU (full chip), 2023-06: CPU 1.12×, GPU 1.32×
- M5 Pro 15 CPU / 16 GPU (binned), 2026-03: CPU 1.16×, GPU 0.74×
- M5 Max 18 CPU / 32 GPU (binned), 2026-03: CPU 1.37×, GPU 1.15×
- M5 Pro 18 CPU / 20 GPU (full chip), 2026-03: CPU 1.32×, GPU 0.8×
16 CPU / 40 GPU (full chip)
- M2 Ultra 24 CPU / 76 GPU (full chip), 2023-06: CPU 0.99×, GPU 1.09×
- M5 Max 18 CPU / 32 GPU (binned), 2026-03: CPU 1.21×, GPU 0.95×
- M3 Ultra 28 CPU / 60 GPU (binned), 2025-03: CPU 1.22×, GPU 1.11×
- M5 Max 18 CPU / 40 GPU (full chip), 2026-03: CPU 1.21×, GPU 1.17×
- M3 Max 16 CPU / 40 GPU (full chip), 2023-11: CPU 0.85×, GPU 0.8×
- M3 Ultra 32 CPU / 80 GPU (full chip), 2025-03: CPU 1.36×, GPU 1.22×