← Back to community benchmarks

gemma-4-26B-A4B-it-ultra-uncensored-heretic

M5 Pro (20c) · 48 GB · 4bit · 2026-05-31
Performance
128k
tokens
967.6
PP tok/s
23.4
TG tok/s
135461
TTFT (ms)
22.4
Peak mem (GB)
Hardware
Chip M5 Pro (20c)
Memory 48 GB
GPU Cores 20
Software
oMLX v0.3.10
macOS macOS 26.4.1
Context 131,072
Performance by Context Length
Context PP tok/s TG tok/s Peak Mem
1k 1,469 81.8 14.0 GB view
4k 1,980 77.7 14.7 GB view
8k 2,077 71.9 15.0 GB view
16k 2,011 63.9 15.5 GB view
32k 1,781 51.3 16.3 GB view
64k 1,393 37.0 18.3 GB view
128k 967.6 23.4 22.4 GB current
Batching Results
Batch Size TG tok/s Speedup
81.8 1.00×
160.9 1.97×
325.4 3.98×