← Back to community benchmarks

Gemma-4-31B-Heretic

M4 Max (40c) · 128 GB · 8bit · 2026-05-04
Performance
8k
tokens
117.5
PP tok/s
6.0
TG tok/s
69725
TTFT (ms)
36.5
Peak mem (GB)
Hardware
Chip M4 Max (40c)
Memory 128 GB
GPU Cores 40
Software
oMLX v0.3.8rc1
macOS macOS 26.4.1
Context 8,192
Performance by Context Length
Context PP tok/s TG tok/s Peak Mem
1k 168.4 7.6 34.5 GB view
4k 121.1 7.4 36.2 GB view
8k 117.5 6.0 36.5 GB current
64k 89.8 3.0 44.5 GB view
Batching Results
Batch Size TG tok/s Speedup
7.6 1.00×
11.3 1.49×
16.8 2.21×