← Back to community benchmarks

Qwen3.5-4B

M5 (10c) · 32 GB · 4bit · 2026-03-10
Performance
4k
tokens
431.6
PP tok/s
45.5
TG tok/s
9491
TTFT (ms)
4.2
Peak mem (GB)
Hardware
Chip M5 (10c)
Memory 32 GB
GPU Cores 10
Software
oMLX v0.2.6
macOS macOS 26.3.1
Context 4,096
Performance by Context Length
Context PP tok/s TG tok/s Peak Mem
1k 423.1 47.8 3.9 GB view
4k 431.6 45.5 4.2 GB current
8k 419.1 42.7 4.6 GB view
16k 394.8 38.1 4.8 GB view
32k 352.6 31.2 5.9 GB view
64k 289.3 22.3 8.0 GB view
Batching Results
Batch Size TG tok/s Speedup
47.8 1.00×
96.4 2.02×
125.1 2.62×