← Back to community benchmarks

Qwen3.6-27B

M5 Max (40c) · 128 GB · 4bit · 2026-04-30
User commentoMLX v0.3.6
Performance
32k
tokens
679.1
PP tok/s
25.7
TG tok/s
48249
TTFT (ms)
21.6
Peak mem (GB)
Hardware
Chip M5 Max (40c)
Memory 128 GB
GPU Cores 40
Software
oMLX v0.3.6
macOS macOS 26.4.1
Context 32,768
Performance by Context Length
Context PP tok/s TG tok/s Peak Mem
1k 790.9 33.1 15.8 GB view
4k 910.8 32.2 17.3 GB view
8k 861.5 31.0 17.9 GB view
16k 746.6 28.6 19.1 GB view
32k 679.1 25.7 21.6 GB current
64k 613.4 20.3 26.7 GB view
128k 457.9 12.1 37.0 GB view
195k 356.8 8.7 47.8 GB view
Batching Results
Batch Size TG tok/s Speedup
33.1 1.00×
59.8 1.81×
77.9 2.35×