← Back to community benchmarks

Qwen3-Coder-30B-A3B-Instruct

M5 Pro (20c) · 48 GB · 8bit · 2026-03-22
Performance
4k
tokens
1,724
PP tok/s
50.9
TG tok/s
2376
TTFT (ms)
31.3
Peak mem (GB)
Hardware
Chip M5 Pro (20c)
Memory 48 GB
GPU Cores 20
Software
oMLX v0.2.20.dev3
macOS macOS 26.3.1
Context 4,096
Performance by Context Length
Context PP tok/s TG tok/s Peak Mem
1k 1,204 57.4 30.8 GB view
4k 1,724 50.9 31.3 GB current
Batching Results
Batch Size TG tok/s Speedup
57.4 1.00×
86.7 1.51×
109.1 1.90×