← Back to community benchmarks

gemma-4-12B-it

M5 Max (40c) · 48 GB · 8bit · 2026-06-04
Performance
128k
tokens
637.2
PP tok/s
18.9
TG tok/s
205693
TTFT (ms)
24.3
Peak mem (GB)
Hardware
Chip M5 Max (40c)
Memory 48 GB
GPU Cores 40
Software
oMLX v0.4.2.dev2
macOS macOS 26.5
Context 131,072
Performance by Context Length
Context PP tok/s TG tok/s Peak Mem
1k 540.3 18.6 18.2 GB view
4k 831.6 13.6 18.6 GB view
8k 910.4 13.4 18.6 GB view
16k 966.2 18.8 18.8 GB view
32k 897.2 24.5 19.2 GB view
64k 795.8 21.2 20.8 GB view
128k 637.2 18.9 24.3 GB current
Batching Results
Batch Size TG tok/s Speedup
18.6 1.00×
32.5 1.75×
52.5 2.82×
120.4 6.47×