← Back to community benchmarks

gemma-4-31b-it

M5 Pro (20c) · 64 GB · 8bit · 2026-04-09
Performance
16k
tokens
331.0
PP tok/s
6.5
TG tok/s
49499
TTFT (ms)
34.8
Peak mem (GB)
Hardware
Chip M5 Pro (20c)
Memory 64 GB
GPU Cores 20
Software
oMLX v0.3.5.dev1
macOS macOS 26.4
Context 16,384
Performance by Context Length
Context PP tok/s TG tok/s Peak Mem
1k 320.3 8.1 31.8 GB view
4k 343.5 7.9 33.6 GB view
8k 336.3 7.4 34.0 GB view
16k 331.0 6.5 34.8 GB current
32k 275.5 4.3 37.1 GB view
64k 234.7 3.3 41.7 GB view
Batching Results
Batch Size TG tok/s Speedup
8.1 1.00×
13.6 1.68×
20.3 2.51×
24.5 3.02×