← Back to community benchmarks

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

M5 Pro (16c) · 48 GB · 8bit · 2026-04-03
Performance
32k
tokens
1,457
PP tok/s
25.7
TG tok/s
22487
TTFT (ms)
27.9
Peak mem (GB)
Hardware
Chip M5 Pro (16c)
Memory 48 GB
GPU Cores 16
Software
oMLX v0.3.2
macOS macOS 26.3
Context 32,768
Performance by Context Length
Context PP tok/s TG tok/s Peak Mem
1k 1,207 43.2 26.1 GB view
4k 1,420 41.2 26.5 GB view
8k 1,628 38.3 26.7 GB view
16k 1,576 33.2 27.1 GB view
32k 1,457 25.7 27.9 GB current
64k 1,168 13.4 29.6 GB view
128k 817.5 6.4 33.1 GB view
195k 567.7 8.0 36.8 GB view
Batching Results
Batch Size TG tok/s Speedup
43.2 1.00×
57.3 1.33×
72.0 1.67×