← Back to community benchmarks

Qwen3-Coder-NextQ9

M3 Ultra (80c) · 512 GB · 8bit · 2026-07-17
Performance
64k
tokens
1,288
PP tok/s
50.4
TG tok/s
50876
TTFT (ms)
90.8
Peak mem (GB)
Hardware
Chip M3 Ultra (80c)
Memory 512 GB
GPU Cores 80
Software
oMLX v0.4.4
macOS macOS 26.5.1
Context 65,536
Performance by Context Length
Context PP tok/s TG tok/s Peak Mem
1k 1,362 63.5 84.8 GB view
4k 1,972 62.4 85.7 GB view
64k 1,288 50.4 90.8 GB current
Batching Results
Batch Size TG tok/s Speedup
63.5 1.00×
100.0 1.57×
157.9 2.49×
231.9 3.65×