← Back to community benchmarks

gemma-4-31b-it

M3 Ultra (80c) · 512 GB · 4bit · 2026-04-06
Performance
195k
tokens
148.8
PP tok/s
3.8
TG tok/s
1344338
TTFT (ms)
47.0
Peak mem (GB)
Hardware
Chip M3 Ultra (80c)
Memory 512 GB
GPU Cores 80
Software
oMLX v0.3.4
macOS macOS 26.2
Context 200,000
Performance by Context Length
Context PP tok/s TG tok/s Peak Mem
1k 317.5 29.8 17.6 GB view
4k 330.5 27.3 19.3 GB view
8k 335.4 24.4 19.7 GB view
16k 324.3 20.2 21.0 GB view
32k 301.6 15.1 22.8 GB view
64k 259.6 9.9 27.5 GB view
128k 196.5 5.8 37.0 GB view
195k 148.8 3.8 47.0 GB current
Batching Results
Batch Size TG tok/s Speedup
29.8 1.00×
49.4 1.66×
65.3 2.19×
75.3 2.53×