← Back to community benchmarks

Qwen3.6-35B-A3B

M5 Pro (20c) · 64 GB · 4bit · 2026-07-21
Performance
1k
tokens
2,079
PP tok/s
95.9
TG tok/s
493
TTFT (ms)
19.3
Peak mem (GB)
Hardware
Chip M5 Pro (20c)
Memory 64 GB
GPU Cores 20
Software
oMLX v0.5.1
macOS macOS 26.5.2
Context 1,024
Performance by Context Length
Context PP tok/s TG tok/s Peak Mem
1k 2,079 95.9 19.3 GB current
4k 2,477 92.9 20.1 GB view
8k 2,397 89.1 20.5 GB view
16k 2,188 82.9 21.4 GB view
32k 1,834 73.4 23.2 GB view
64k 1,317 59.2 25.8 GB view
128k 834.0 43.8 32.1 GB view
195k 573.1 32.9 39.2 GB view
Batching Results
Batch Size TG tok/s Speedup
95.9 1.00×
153.7 1.60×
214.6 2.24×
436.1 4.55×