← Back to community benchmarks

Qwen3.6-35B-A3B

M5 Pro (20c) · 64 GB · 4bit · 2026-04-26
Performance
16k
tokens
1,671
PP tok/s
92.2
TG tok/s
9806
TTFT (ms)
38.8
Peak mem (GB)
Hardware
Chip M5 Pro (20c)
Memory 64 GB
GPU Cores 20
Software
oMLX v0.3.6
macOS macOS 26.4.1
Context 16,384
Performance by Context Length
Context PP tok/s TG tok/s Peak Mem
1k 1,456 102.6 36.4 GB view
4k 2,167 97.5 37.2 GB view
8k 2,289 92.3 37.5 GB view
16k 1,671 92.2 38.8 GB current
32k 1,456 89.4 39.5 GB view
64k 1,089 72.0 42.2 GB view
128k 732.9 65.8 48.2 GB view
Batching Results
Batch Size TG tok/s Speedup
102.6 1.00×
79.2 0.77×
191.6 1.87×
231.0 2.25×