← Back to community benchmarks

Qwen3.5-27B

M3 Ultra (80c) · 512 GB · bf16 · 2026-03-22
Performance
16k
tokens
404.9
PP tok/s
11.5
TG tok/s
40464
TTFT (ms)
54.2
Peak mem (GB)
Hardware
Chip M3 Ultra (80c)
Memory 512 GB
GPU Cores 80
Software
oMLX v0.2.11
macOS macOS 15.3
Context 16,384
Performance by Context Length
Context PP tok/s TG tok/s Peak Mem
1k 384.8 12.5 52.5 GB view
4k 416.1 12.2 52.8 GB view
8k 417.3 11.9 53.3 GB view
16k 404.9 11.5 54.2 GB current
Batching Results
Batch Size TG tok/s Speedup
12.5 1.00×
13.7 1.10×
27.1 2.17×
51.7 4.14×