← Back to community benchmarks

Qwen3-Coder-30B-A3B-Instruct5bit

M4 Max (40c) · 64 GB · 5bit · 2026-03-10
Performance
16k
tokens
730.7
PP tok/s
43.1
TG tok/s
22421
TTFT (ms)
21.7
Peak mem (GB)
Hardware
Chip M4 Max (40c)
Memory 64 GB
GPU Cores 40
Software
oMLX v0.2.7
macOS macOS 26.3.1
Context 16,384
Performance by Context Length
Context PP tok/s TG tok/s Peak Mem
1k 1,286 90.7 20.1 GB view
4k 1,337 75.7 20.6 GB view
8k 1,116 64.3 20.9 GB view
16k 730.7 43.1 21.7 GB current
32k 398.7 26.3 23.2 GB view
64k 190.1 15.8 26.3 GB view
Batching Results
Batch Size TG tok/s Speedup
90.7 1.00×
106.6 1.18×
128.0 1.41×
152.1 1.68×