← Back to community benchmarks

Qwen3.5-27B

M3 Max (40c) · 48 GB · 4bit · 2026-03-12
Performance
16k
tokens
176.5
PP tok/s
15.5
TG tok/s
92846
TTFT (ms)
18.3
Peak mem (GB)
Hardware
Chip M3 Max (40c)
Memory 48 GB
GPU Cores 40
Software
oMLX v0.2.8
macOS macOS 26.3.1
Context 16,384
Performance by Context Length
Context PP tok/s TG tok/s Peak Mem
1k 172.1 17.9 16.6 GB view
4k 188.6 19.1 17.0 GB view
8k 186.9 17.5 17.4 GB view
16k 176.5 15.5 18.3 GB current
Batching Results
Batch Size TG tok/s Speedup
17.9 1.00×
28.4 1.59×
35.6 1.99×