← Back to community benchmarks

community--Qwen3-30B-A3B-Instruct-2507

M5 Max (40c) · 128 GB · 6bit · 2026-07-24
Performance
4k
tokens
4,334
PP tok/s
96.3
TG tok/s
945
TTFT (ms)
24.2
Peak mem (GB)
Hardware
Chip M5 Max (40c)
Memory 128 GB
GPU Cores 40
Software
oMLX v0.5.3
macOS macOS 26.3
Context 4,096
Performance by Context Length
Context PP tok/s TG tok/s Peak Mem
1k 3,797 101.4 23.7 GB view
4k 4,334 96.3 24.2 GB current
8k 3,650 82.8 24.5 GB view
16k 2,661 72.5 25.3 GB view
32k 1,775 52.8 26.7 GB view
64k 1,091 30.4 29.7 GB view
128k 437.7 12.2 35.8 GB view
195k 265.6 8.8 42.5 GB view
Batching Results
Batch Size TG tok/s Speedup
101.4 1.00×
163.5 1.61×
322.2 3.18×
438.4 4.32×