← Back to community benchmarks

Qwen3.5-27B

M5 Max (40c) · 128 GB · 4bit · 2026-04-07
Performance
195k
tokens
343.3
PP tok/s
7.9
TG tok/s
582499
TTFT (ms)
47.8
Peak mem (GB)
Hardware
Chip M5 Max (40c)
Memory 128 GB
GPU Cores 40
Software
oMLX v0.3.4
macOS macOS 26.4
Context 200,000
Performance by Context Length
Context PP tok/s TG tok/s Peak Mem
1k 781.5 32.7 15.8 GB view
4k 880.0 31.6 17.3 GB view
32k 681.7 25.3 21.6 GB view
64k 578.6 20.1 26.7 GB view
128k 434.9 13.7 37.0 GB view
195k 343.3 7.9 47.8 GB current
Batching Results
Batch Size TG tok/s Speedup
32.7 1.00×
57.4 1.76×
74.8 2.29×
82.6 2.53×