← Back to community benchmarks

Qwen3.5-4B

M1 (8c) · 8 GB · 4bit · 2026-04-12
Performance
8k
tokens
141.6
PP tok/s
20.0
TG tok/s
57862
TTFT (ms)
4.3
Peak mem (GB)
Hardware
Chip M1 (8c)
Memory 8 GB
GPU Cores 8
Software
oMLX v0.3.4
macOS macOS 26.4
Context 8,192
Performance by Context Length
Context PP tok/s TG tok/s Peak Mem
1k 140.4 22.7 3.3 GB view
4k 143.6 21.6 3.9 GB view
8k 141.6 20.0 4.3 GB current
16k 135.7 17.4 5.0 GB view
Batching Results
Batch Size TG tok/s Speedup
22.7 1.00×
30.4 1.34×
33.8 1.49×
29.4 1.30×