← Back to community benchmarks

Qwen3.5-27B-Claude-4.6-Opus-Distilled

M3 Ultra (80c) · 96 GB · 4bit · 2026-03-13
Performance
16k
tokens
406.8
PP tok/s
31.3
TG tok/s
40271
TTFT (ms)
17.5
Peak mem (GB)
Hardware
Chip M3 Ultra (80c)
Memory 96 GB
GPU Cores 80
Software
oMLX v0.2.10
macOS macOS 15.3
Context 16,384
Performance by Context Length
Context PP tok/s TG tok/s Peak Mem
1k 389.1 39.1 15.8 GB view
4k 417.8 36.5 16.1 GB view
8k 417.8 34.9 16.6 GB view
16k 406.8 31.3 17.5 GB current
32k 381.8 26.3 19.2 GB view
64k 333.8 18.1 22.8 GB view
Batching Results
Batch Size TG tok/s Speedup
39.1 1.00×
69.9 1.79×
90.5 2.31×
104.3 2.67×