← Back to community benchmarks

Qwen3.5-27B

M5 Max (40c) · 128 GB · 8bit · 2026-03-18
Performance
195k
tokens
343.9
PP tok/s
5.9
TG tok/s
581546
TTFT (ms)
61.3
Peak mem (GB)
Hardware
Chip M5 Max (40c)
Memory 128 GB
GPU Cores 40
Software
oMLX v0.2.17
macOS macOS 26.3.1
Context 200,000
Performance by Context Length
Context PP tok/s TG tok/s Peak Mem
1k 808.0 18.0 29.2 GB view
4k 851.8 17.6 30.8 GB view
8k 814.0 17.2 31.4 GB view
16k 712.2 13.6 32.7 GB view
32k 666.4 13.5 35.2 GB view
64k 579.3 10.7 40.3 GB view
128k 431.1 7.8 50.6 GB view
195k 343.9 5.9 61.3 GB current
Batching Results
Batch Size TG tok/s Speedup
18.0 1.00×
34.6 1.92×
48.9 2.72×
57.5 3.19×