2benchesM4
A18 · M4 · 2024 CPU family · 3 nm · First available 2024-05-15
Every configuration
| Configuration | CPU core layout | Bandwidth | GB7 CPU | GB7 Metal | Prefill Q4_0 tok/s | Decode Q4_0 tok/s |
|---|---|---|---|---|---|---|
| 8 CPU / 8 GPU (binned) | 4P + 4E | 120 GB/s | 13,419 | 47,224 | — | — |
| 10 CPU / 8 GPU (binned) | 4P + 6E | 120 GB/s | 15,044 | 47,547 | — | — |
| 9 CPU / 10 GPU (binned) | 3P + 6E | 120 GB/s | — | — | — | — |
| 10 CPU / 10 GPU (full chip) | 4P + 6E | 120 GB/s | 15,184 | 53,143 | — | — |
All benchmarks
| Benchmark | Median |
|---|---|
| Geekbench 7 · Multi-core | 15,044 (13,419–15,497, 9 sources) |
| Geekbench 7 · Metal GPU | 52,783 (47,224–54,095, 7 sources) |
| Geekbench AI · GPU · Single | 9,184 (9,183–9,186, 2 sources) |
| Geekbench AI · GPU · Half | 10,847 (10,846–10,847, 2 sources) |
| Geekbench AI · GPU · Quantized | 9,503 (9,479–9,526, 2 sources) |
| Geekbench AI · Neural Engine · Single | 4,580 (4,556–4,605, 2 sources) |
| Geekbench AI · Neural Engine · Half | 33,426 (33,138–33,714, 2 sources) |
| Geekbench AI · Neural Engine · Quantized | 46,321 (45,882–46,761, 2 sources) |
| Prefill 512 · F16 | 230.2 (230.2–230.2, 1 sources) |
| Decode 128 · F16 | 7.4 (7.4–7.4, 1 sources) |
| Prefill 512 · Q8_0 | 223.6 (223.6–223.6, 1 sources) |
| Decode 128 · Q8_0 | 13.5 (13.5–13.5, 1 sources) |
| Prefill 512 · Q4_0 | 221.3 (221.3–221.3, 1 sources) |
| Decode 128 · Q4_0 | 24.1 (24.1–24.1, 1 sources) |
Similar performance
Nearest measured configurations by Geekbench 7 CPU and Metal. Ratios above 1 are faster.
8 CPU / 8 GPU (binned)
- M3 8 CPU / 10 GPU (full chip), 2023-11: CPU 0.87×, GPU 1.01×
- M3 8 CPU / 8 GPU (binned), 2023-11: CPU 0.87×, GPU 0.91×
- M2 8 CPU / 10 GPU (full chip), 2022-06: CPU 0.72×, GPU 0.91×
- M5 10 CPU / 8 GPU (binned), 2025-10: CPU 1.27×, GPU 1.39×
- M3 Pro 11 CPU / 14 GPU (binned), 2023-11: CPU 1.14×, GPU 1.49×
- M2 Pro 10 CPU / 16 GPU (binned), 2023-01: CPU 1.02×, GPU 1.57×
10 CPU / 8 GPU (binned)
- M3 8 CPU / 10 GPU (full chip), 2023-11: CPU 0.77×, GPU 1.01×
- M3 8 CPU / 8 GPU (binned), 2023-11: CPU 0.78×, GPU 0.9×
- M5 10 CPU / 8 GPU (binned), 2025-10: CPU 1.13×, GPU 1.38×
- M3 Pro 11 CPU / 14 GPU (binned), 2023-11: CPU 1.02×, GPU 1.48×
- M5 10 CPU / 10 GPU (full chip), 2025-10: CPU 1.14×, GPU 1.49×
- M2 8 CPU / 10 GPU (full chip), 2022-06: CPU 0.64×, GPU 0.91×
10 CPU / 10 GPU (full chip)
- M5 10 CPU / 8 GPU (binned), 2025-10: CPU 1.12×, GPU 1.24×
- M3 Pro 11 CPU / 14 GPU (binned), 2023-11: CPU 1.01×, GPU 1.32×
- M3 8 CPU / 10 GPU (full chip), 2023-11: CPU 0.76×, GPU 0.9×
- M5 10 CPU / 10 GPU (full chip), 2025-10: CPU 1.13×, GPU 1.34×
- M3 8 CPU / 8 GPU (binned), 2023-11: CPU 0.77×, GPU 0.81×
- M2 Pro 10 CPU / 16 GPU (binned), 2023-01: CPU 0.9×, GPU 1.39×