← Back to community benchmarks

gemma-4-26b-a4b-it-mxfp8

M4 Max (40c) · 128 GB · 8bit · 2026-04-03
Performance
128k
tokens
685.3
PP tok/s
1.2
TG tok/s
191259
TTFT (ms)
33.3
Peak mem (GB)
Hardware
Chip M4 Max (40c)
Memory 128 GB
GPU Cores 40
Software
oMLX v0.3.2
macOS macOS 15.7.5
Context 131,072
Performance by Context Length
Context PP tok/s TG tok/s Peak Mem
1k 1,358 57.8 26.1 GB view
4k 1,365 55.7 26.6 GB view
8k 1,327 51.2 26.8 GB view
16k 1,281 37.9 27.2 GB view
32k 1,153 24.9 28.0 GB view
64k 951.6 14.7 29.7 GB view
128k 685.3 1.2 33.3 GB current
Batching Results
Batch Size TG tok/s Speedup
57.8 1.00×
79.9 1.38×
108.6 1.88×