← Back to community benchmarks

gemma-4-26b-a4b-it

M3 Ultra (60c) · 256 GB · 8bit · 2026-04-03
Performance
16k
tokens
1,655
PP tok/s
43.2
TG tok/s
9899
TTFT (ms)
28.5
Peak mem (GB)
Hardware
Chip M3 Ultra (60c)
Memory 256 GB
GPU Cores 60
Software
oMLX v0.3.2
macOS macOS 26.3
Context 16,384
Performance by Context Length
Context PP tok/s TG tok/s Peak Mem
1k 1,718 56.7 27.5 GB view
4k 1,766 52.7 27.9 GB view
8k 1,743 48.6 28.0 GB view
16k 1,655 43.2 28.5 GB current
32k 1,497 33.1 29.3 GB view
64k 1,226 25.1 31.0 GB view
128k 881.0 14.6 34.5 GB view
195k 662.2 10.0 38.3 GB view
Batching Results
Batch Size TG tok/s Speedup
56.7 1.00×
72.7 1.28×
121.4 2.14×
185.1 3.26×