← Back to community benchmarks

Qwen3.5-4B

M5 (10c) · 32 GB · 4bit · 2026-03-12
Performance
16k
tokens
386.7
PP tok/s
34.0
TG tok/s
42372
TTFT (ms)
4.8
Peak mem (GB)
Hardware
Chip M5 (10c)
Memory 32 GB
GPU Cores 10
Software
oMLX v0.2.7
macOS macOS 26.3.1
Context 16,384
Performance by Context Length
Context PP tok/s TG tok/s Peak Mem
1k 425.9 46.4 3.9 GB view
4k 423.9 45.2 4.2 GB view
8k 411.7 39.9 4.6 GB view
16k 386.7 34.0 4.8 GB current
Batching Results
Batch Size TG tok/s Speedup
46.4 1.00×
96.4 2.08×
130.5 2.81×