← Back to community benchmarks

gemma-4-26B-A4B-it5bit

M2 Ultra (60c) · 64 GB · 5bit · 2026-04-23
Performance
16k
tokens
884.4
PP tok/s
34.8
TG tok/s
18525
TTFT (ms)
18.3
Peak mem (GB)
Hardware
Chip M2 Ultra (60c)
Memory 64 GB
GPU Cores 60
Software
oMLX v0.3.7
macOS macOS 15.7.4
Context 16,384
Performance by Context Length
Context PP tok/s TG tok/s Peak Mem
1k 743.4 61.8 17.1 GB view
4k 867.9 56.7 17.7 GB view
8k 895.8 50.4 17.9 GB view
16k 884.4 34.8 18.3 GB current
32k 831.7 22.0 19.0 GB view
64k 719.6 9.8 20.7 GB view
Batching Results
Batch Size TG tok/s Speedup
61.8 1.00×
87.5 1.42×
121.0 1.96×
145.0 2.35×