← Back to community benchmarks

gemma-4-26b

M3 Pro (18c) · 36 GB · 4bit · 2026-05-06
Performance
8k
tokens
518.3
PP tok/s
38.6
TG tok/s
15805
TTFT (ms)
15.1
Peak mem (GB)
Hardware
Chip M3 Pro (18c)
Memory 36 GB
GPU Cores 18
Software
oMLX v0.3.8
macOS macOS 26.4.1
Context 8,192
Performance by Context Length
Context PP tok/s TG tok/s Peak Mem
1k 460.6 42.1 14.3 GB view
4k 536.4 40.4 14.9 GB view
8k 518.3 38.6 15.1 GB current
16k 488.4 35.4 15.2 GB view
32k 433.3 20.9 16.1 GB view
64k 266.1 8.6 17.7 GB view
Batching Results
Batch Size TG tok/s Speedup
42.1 1.00×
61.9 1.47×
72.3 1.72×
91.4 2.17×