← Back to community benchmarks

Qwen3.5-27B-mxfp8

M5 Pro (20c) · 64 GB · 8bit · 2026-03-13
Performance
1k
tokens
410.1
PP tok/s
10.1
TG tok/s
2497
TTFT (ms)
28.4
Peak mem (GB)
Hardware
Chip M5 Pro (20c)
Memory 64 GB
GPU Cores 20
Software
oMLX v0.2.9
macOS macOS 26.3.1
Context 1,024
Performance by Context Length
Context PP tok/s TG tok/s Peak Mem
1k 410.1 10.1 28.4 GB current
4k 434.5 10.0 28.7 GB view
8k 430.7 9.7 29.2 GB view
16k 408.5 9.3 30.0 GB view
32k 380.2 8.6 31.8 GB view
64k 321.4 7.4 35.4 GB view
Batching Results
Batch Size TG tok/s Speedup
10.1 1.00×
18.5 1.83×
24.2 2.40×
25.3 2.50×