← Back to community benchmarks

Qwen3.6-27B-UD

M3 Ultra (80c) · 512 GB · 4bit · 2026-04-29
Performance
4k
tokens
418.9
PP tok/s
23.8
TG tok/s
9778
TTFT (ms)
26.6
Peak mem (GB)
Hardware
Chip M3 Ultra (80c)
Memory 512 GB
GPU Cores 80
Software
oMLX v0.3.8rc1
macOS macOS 26.3.1
Context 4,096
Performance by Context Length
Context PP tok/s TG tok/s Peak Mem
1k 374.9 24.5 25.2 GB view
4k 418.9 23.8 26.6 GB current
8k 420.0 23.1 27.2 GB view
Batching Results
Batch Size TG tok/s Speedup
24.5 1.00×
30.4 1.24×
50.7 2.07×
75.9 3.10×