← Back to community benchmarks

Qwen3.5-9B

M5 (10c) · 24 GB · 4bit · 2026-03-10
Performance
32k
tokens
207.2
PP tok/s
19.5
TG tok/s
158137
TTFT (ms)
8.5
Peak mem (GB)
Hardware
Chip M5 (10c)
Memory 24 GB
GPU Cores 10
Software
oMLX v0.2.7
macOS macOS 26.3
Context 32,768
Performance by Context Length
Context PP tok/s TG tok/s Peak Mem
1k 224.4 24.6 6.5 GB view
4k 231.9 25.0 6.8 GB view
8k 228.0 23.5 7.0 GB view
16k 216.8 22.1 7.5 GB view
32k 207.2 19.5 8.5 GB current
64k 184.0 12.2 10.6 GB view
Batching Results
Batch Size TG tok/s Speedup
24.6 1.00×
74.4 3.02×