← Back to community benchmarks

gemma-4-26b-a4b-mxfp8

M3 Max (40c) · 128 GB · 8bit · 2026-05-11
Performance
16k
tokens
1,093
PP tok/s
55.1
TG tok/s
14993
TTFT (ms)
26.9
Peak mem (GB)
Hardware
Chip M3 Max (40c)
Memory 128 GB
GPU Cores 40
Software
oMLX v0.3.8
macOS macOS 15.7.4
Context 16,384
Performance by Context Length
Context PP tok/s TG tok/s Peak Mem
1k 860.5 61.3 25.0 GB view
4k 1,107 60.3 25.9 GB view
8k 1,139 58.3 26.3 GB view
16k 1,093 55.1 26.9 GB current
32k 985.1 49.5 28.2 GB view
64k 803.9 41.9 30.9 GB view
128k 583.5 31.3 36.5 GB view
195k 438.7 12.8 48.4 GB view
Batching Results
Batch Size TG tok/s Speedup
61.3 1.00×
89.5 1.46×
110.4 1.80×