← Back to community benchmarks

gemma4H-31B-oQ8e

M3 Ultra (80c) · 512 GB · 8bit · 2026-07-17
Performance
16k
tokens
312.0
PP tok/s
16.3
TG tok/s
52514
TTFT (ms)
41.2
Peak mem (GB)
Hardware
Chip M3 Ultra (80c)
Memory 512 GB
GPU Cores 80
Software
oMLX v0.5.1
macOS macOS 26.5.2
Context 16,384
Performance by Context Length
Context PP tok/s TG tok/s Peak Mem
1k 348.7 17.9 31.8 GB view
4k 339.2 17.5 33.6 GB view
8k 326.1 17.1 37.5 GB view
16k 312.0 16.3 41.2 GB current
32k 289.3 14.9 49.6 GB view
64k 249.5 12.6 63.8 GB view
128k 190.2 9.5 76.4 GB view