← Back to community benchmarks

Qwen3.5-27B

M4 Max (40c) · 64 GB · 4bit · 2026-03-13
Performance
16k
tokens
167.2
PP tok/s
12.2
TG tok/s
97972
TTFT (ms)
18.3
Peak mem (GB)
Hardware
Chip M4 Max (40c)
Memory 64 GB
GPU Cores 40
Software
oMLX v0.2.6
macOS macOS 26.3
Context 16,384
Performance by Context Length
Context PP tok/s TG tok/s Peak Mem
1k 247.4 28.3 16.6 GB view
4k 204.0 24.8 17.0 GB view
8k 193.5 24.2 17.4 GB view
16k 167.2 12.2 18.3 GB current
Batching Results
Batch Size TG tok/s Speedup
28.3 1.00×
20.6 0.73×
35.9 1.27×