← Back to community benchmarks

Qwen3.5-9B

M5 Pro (20c) · 24 GB · 4bit · 2026-03-18
Performance
4k
tokens
1,632
PP tok/s
54.5
TG tok/s
2510
TTFT (ms)
7.4
Peak mem (GB)
Hardware
Chip M5 Pro (20c)
Memory 24 GB
GPU Cores 20
Software
oMLX v0.2.18
macOS macOS 26.3.1
Context 4,096
Performance by Context Length
Context PP tok/s TG tok/s Peak Mem
1k 1,606 56.4 6.7 GB view
4k 1,632 54.5 7.4 GB current
8k 1,635 53.4 7.8 GB view
16k 1,551 49.3 8.6 GB view
32k 1,393 42.2 10.1 GB view
64k 1,158 32.8 13.2 GB view
Batching Results
Batch Size TG tok/s Speedup
56.4 1.00×
104.8 1.86×
140.3 2.49×
149.3 2.65×