← Back to community benchmarks

Qwen3.5-9B

M5 (10c) · 32 GB · 4bit · 2026-03-11
Performance
8k
tokens
233.3
PP tok/s
24.8
TG tok/s
35117
TTFT (ms)
7.0
Peak mem (GB)
Hardware
Chip M5 (10c)
Memory 32 GB
GPU Cores 10
Software
oMLX v0.2.7
macOS macOS 26.3.1
Context 8,192
Performance by Context Length
Context PP tok/s TG tok/s Peak Mem
1k 233.7 26.4 6.5 GB view
4k 236.8 26.1 6.8 GB view
8k 233.3 24.8 7.0 GB current
Batching Results
Batch Size TG tok/s Speedup
26.4 1.00×
56.3 2.13×
72.4 2.74×