← Back to community benchmarks

gemma-4-26B-A4B-it-qat-5bit

M5 Pro (20c) · 48 GB · 5bit · 2026-07-19
Performance
1k
tokens
1,918
PP tok/s
65.0
TG tok/s
534
TTFT (ms)
17.1
Peak mem (GB)
Hardware
Chip M5 Pro (20c)
Memory 48 GB
GPU Cores 20
Software
oMLX v0.5.2.dev1
macOS macOS 26.5.2
Context 1,024
Performance by Context Length
Context PP tok/s TG tok/s Peak Mem
1k 1,918 65.0 17.1 GB current
4k 2,036 58.3 17.6 GB view
8k 1,914 58.9 18.0 GB view
16k 1,861 53.4 18.4 GB view
32k 1,625 44.3 19.4 GB view
64k 1,285 31.8 21.4 GB view
Batching Results
Batch Size TG tok/s Speedup
65.0 1.00×
102.5 1.58×
137.1 2.11×