← Back to community benchmarks

Qwen3.6-27B-oQ8-mtp

M3 Ultra (60c) · 96 GB · 8bit · 2026-05-07
Performance
32k
tokens
292.2
PP tok/s
27.4
TG tok/s
112147
TTFT (ms)
35.8
Peak mem (GB)
Hardware
Chip M3 Ultra (60c)
Memory 96 GB
GPU Cores 60
Software
oMLX v0.3.9.dev1
macOS macOS 26.4.1
Context 32,768
Performance by Context Length
Context PP tok/s TG tok/s Peak Mem
1k 299.4 33.1 28.8 GB view
4k 319.7 32.6 30.3 GB view
8k 318.5 31.6 31.3 GB view
16k 310.2 28.9 32.8 GB view
32k 292.2 27.4 35.8 GB current
64k 257.1 23.0 41.8 GB view
128k 203.5 18.6 54.2 GB view
195k 166.4 15.5 68.9 GB view
Batching Results
Batch Size TG tok/s Speedup
33.1 1.00×
39.0 1.18×
50.1 1.51×
58.2 1.76×