← Back to community benchmarks

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

M3 Ultra (60c) · 256 GB · 8bit · 2026-05-08
Performance
32k
tokens
1,468
PP tok/s
60.2
TG tok/s
22316
TTFT (ms)
27.3
Peak mem (GB)
Hardware
Chip M3 Ultra (60c)
Memory 256 GB
GPU Cores 60
Software
oMLX v0.3.9.dev1
macOS macOS 26.4.1
Context 32,768
Performance by Context Length
Context PP tok/s TG tok/s Peak Mem
1k 1,007 80.5 25.1 GB view
4k 1,470 78.4 25.7 GB view
8k 1,682 75.5 25.9 GB view
16k 1,612 69.9 26.3 GB view
32k 1,468 60.2 27.3 GB current
64k 1,215 45.7 29.3 GB view
128k 879.8 31.2 33.5 GB view
195k 665.7 19.9 38.4 GB view
Batching Results
Batch Size TG tok/s Speedup
80.5 1.00×
116.7 1.45×
148.8 1.85×
295.8 3.67×