← Back to community benchmarks

Qwen3.5-397B-A17B

M3 Ultra (80c) · 512 GB · 6bit · 2026-03-08
Performance
16k
tokens
462.0
PP tok/s
23.6
TG tok/s
35462
TTFT (ms)
304.0
Peak mem (GB)
Hardware
Chip M3 Ultra (80c)
Memory 512 GB
GPU Cores 80
Software
oMLX v0.2.6
macOS macOS 15.5
Context 16,384
Performance by Context Length
Context PP tok/s TG tok/s Peak Mem
1k 454.0 28.5 302.7 GB view
4k 487.1 27.2 302.9 GB view
8k 484.9 26.0 303.3 GB view
16k 462.0 23.6 304.0 GB current
32k 422.5 20.2 305.5 GB view
Batching Results
Batch Size TG tok/s Speedup
28.5 1.00×
44.3 1.55×
57.9 2.03×
81.5 2.86×