← Back to community benchmarks

Qwen3.6-27B4bit-MTP

M4 (10c) · 32 GB · 4bit · 2026-05-19
Performance
1k
tokens
62.6
PP tok/s
10.6
TG tok/s
16366
TTFT (ms)
16.0
Peak mem (GB)
Hardware
Chip M4 (10c)
Memory 32 GB
GPU Cores 10
Software
oMLX v0.3.9.dev2
macOS macOS 26.3.1
Context 1,024
Performance by Context Length
Context PP tok/s TG tok/s Peak Mem
1k 62.6 10.6 16.0 GB current
4k 63.0 11.1 17.4 GB view
Batching Results
Batch Size TG tok/s Speedup
10.6 1.00×
13.2 1.25×
16.4 1.55×