← Back to community benchmarks

Qwen3.5-4B

M1 Max (32c) · 64 GB · 8bit · 2026-05-02
Performance
8k
tokens
570.9
PP tok/s
56.6
TG tok/s
14350
TTFT (ms)
6.3
Peak mem (GB)
Hardware
Chip M1 Max (32c)
Memory 64 GB
GPU Cores 32
Software
oMLX v0.3.8
macOS macOS 26.4.1
Context 8,192
Performance by Context Length
Context PP tok/s TG tok/s Peak Mem
1k 501.9 61.3 5.3 GB view
4k 571.1 59.2 6.0 GB view
8k 570.9 56.6 6.3 GB current
Batching Results
Batch Size TG tok/s Speedup
61.3 1.00×
104.9 1.71×
123.0 2.01×
113.8 1.86×