← Back to community benchmarks

Qwen3.5-9B-MTPLX-Optimized-Speed

M5 (10c) · 24 GB · 6bit · 2026-07-13
Performance
64k
tokens
408.5
PP tok/s
12.9
TG tok/s
160439
TTFT (ms)
14.7
Peak mem (GB)
Hardware
Chip M5 (10c)
Memory 24 GB
GPU Cores 10
Software
oMLX v0.5.1
macOS macOS 26.5.2
Context 65,536
Performance by Context Length
Context PP tok/s TG tok/s Peak Mem
1k 661.9 18.1 7.8 GB view
4k 671.3 17.7 8.3 GB view
8k 657.1 17.4 8.8 GB view
16k 615.0 16.8 9.9 GB view
32k 517.4 15.3 11.7 GB view
64k 408.5 12.9 14.7 GB current
Batching Results
Batch Size TG tok/s Speedup
18.1 1.00×
34.6 1.91×
55.1 3.04×