← Back to community benchmarks

gemma-4-31b-it

M2 Ultra (60c) · 128 GB · 8bit · 2026-07-19
Performance
195k
tokens
75.0
PP tok/s
6.0
TG tok/s
2667560
TTFT (ms)
76.5
Peak mem (GB)
Hardware
Chip M2 Ultra (60c)
Memory 128 GB
GPU Cores 60
Software
oMLX v0.5.1
macOS macOS 26.5
Context 200,000
Performance by Context Length
Context PP tok/s TG tok/s Peak Mem
1k 152.3 16.9 31.8 GB view
4k 149.2 16.5 33.6 GB view
8k 145.7 15.9 36.5 GB view
16k 141.6 14.8 40.6 GB view
32k 134.8 13.1 40.9 GB view
64k 119.7 10.7 46.8 GB view
128k 92.2 7.8 59.7 GB view
195k 75.0 6.0 76.5 GB current
Batching Results
Batch Size TG tok/s Speedup
16.9 1.00×
26.1 1.54×
31.8 1.88×
36.4 2.15×