← Back to community benchmarks

Qwen3.5-27BMTP

M3 Ultra (60c) · 96 GB · 4bit · 2026-05-19
Performance
1k
tokens
304.3
PP tok/s
50.6
TG tok/s
3365
TTFT (ms)
16.1
Peak mem (GB)
Hardware
Chip M3 Ultra (60c)
Memory 96 GB
GPU Cores 60
Software
oMLX v0.3.9rc1
macOS macOS 26.3.1
Context 1,024
Performance by Context Length
Context PP tok/s TG tok/s Peak Mem
1k 304.3 50.6 16.1 GB current
4k 321.5 45.7 17.6 GB view
8k 320.3 43.0 18.6 GB view
Batching Results
Batch Size TG tok/s Speedup
50.6 1.00×
62.6 1.24×
72.1 1.42×