← Back to community benchmarks

Qwen3.6-27B-oQ4-mtp

M5 (10c) · 24 GB · 4bit · 2026-05-19
Performance
8k
tokens
219.0
PP tok/s
7.4
TG tok/s
37399
TTFT (ms)
18.8
Peak mem (GB)
Hardware
Chip M5 (10c)
Memory 24 GB
GPU Cores 10
Software
oMLX v0.3.9rc1
macOS macOS 26.4
Context 8,192
Performance by Context Length
Context PP tok/s TG tok/s Peak Mem
1k 218.8 7.7 16.4 GB view
4k 229.3 7.6 17.8 GB view
8k 219.0 7.4 18.8 GB current
Batching Results
Batch Size TG tok/s Speedup
7.7 1.00×
15.6 2.03×
20.8 2.70×