2benchesM5 Pro
A19 · M5 · 2025 CPU family · 3 nm · First available 2026-03-11
Every configuration
| Configuration | CPU core layout | Bandwidth | GB7 CPU | GB7 Metal | Prefill Q4_0 tok/s | Decode Q4_0 tok/s |
|---|---|---|---|---|---|---|
| 15 CPU / 16 GPU (binned) | 5S + 10P | 307 GB/s | 29,587 | 124,511 | — | — |
| 18 CPU / 20 GPU (full chip) | 6S + 12P | 307 GB/s | 33,532 | 135,752 | 1,621 | 66.3 |
All benchmarks
| Benchmark | Median |
|---|---|
| Geekbench 7 · Multi-core | 33,467 (29,587–33,597, 3 sources) |
| Geekbench 7 · Metal GPU | 135,318 (124,511–136,185, 3 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 | 1,589 (1,589–1,589, 1 sources) |
| Decode 128 · F16 | 21.6 (21.6–21.6, 1 sources) |
| Prefill 512 · Q8_0 | 1,554 (1,554–1,554, 1 sources) |
| Decode 128 · Q8_0 | 38.9 (38.9–38.9, 1 sources) |
| Prefill 512 · Q4_0 | 1,621 (1,621–1,621, 1 sources) |
| Decode 128 · Q4_0 | 66.3 (66.3–66.3, 1 sources) |
Similar performance
Nearest measured configurations by Geekbench 7 CPU and Metal. Ratios above 1 are faster.
15 CPU / 16 GPU (binned)
- M4 Pro 14 CPU / 20 GPU (full chip), 2024-11: CPU 0.84×, GPU 0.92×
- M3 Max 14 CPU / 30 GPU (binned), 2023-11: CPU 0.74×, GPU 1.06×
- M3 Max 16 CPU / 40 GPU (full chip), 2023-11: CPU 0.83×, GPU 1.31×
- M4 Max 14 CPU / 32 GPU (binned), 2024-11: CPU 0.86×, GPU 1.35×
- M4 Pro 12 CPU / 16 GPU (binned), 2024-11: CPU 0.73×, GPU 0.82×
- M5 Max 18 CPU / 32 GPU (binned), 2026-03: CPU 1.18×, GPU 1.56×
18 CPU / 20 GPU (full chip)
- M4 Pro 14 CPU / 20 GPU (full chip), 2024-11: CPU 0.74×, GPU 0.85×
- M4 Max 14 CPU / 32 GPU (binned), 2024-11: CPU 0.76×, GPU 1.24×
- M5 Max 18 CPU / 32 GPU (binned), 2026-03: CPU 1.04×, GPU 1.43×
- M3 Max 16 CPU / 40 GPU (full chip), 2023-11: CPU 0.73×, GPU 1.2×
- M3 Max 14 CPU / 30 GPU (binned), 2023-11: CPU 0.66×, GPU 0.97×
- M4 Max 16 CPU / 40 GPU (full chip), 2024-11: CPU 0.86×, GPU 1.5×