← Back to community benchmarks

Qwen3.5-35B-A3B

M5 Pro (20c) · 64 GB · 4bit · 2026-04-18
Performance
4k
tokens
2,150
PP tok/s
98.9
TG tok/s
1905
TTFT (ms)
20.0
Peak mem (GB)
Hardware
Chip M5 Pro (20c)
Memory 64 GB
GPU Cores 20
Software
oMLX v0.3.6
macOS macOS 26.4.1
Context 4,096
Performance by Context Length
Context PP tok/s TG tok/s Peak Mem
1k 1,341 101.8 19.3 GB view
4k 2,150 98.9 20.0 GB current
8k 2,241 85.9 20.4 GB view
16k 2,144 79.3 21.0 GB view
32k 1,816 59.2 22.3 GB view
64k 1,332 41.0 25.0 GB view
128k 833.0 26.3 30.4 GB view
195k 602.6 19.3 36.1 GB view
Batching Results
Batch Size TG tok/s Speedup
101.8 1.00×
150.8 1.48×
196.9 1.93×
240.3 2.36×