← Back to community benchmarks

Qwen3.5-9B-Claude-4.6-HighIQ-INSTRUCT-HERETIC-UNCENSOREDmxfp8

M5 (8c) · 16 GB · 8bit · 2026-04-06
Performance
8k
tokens
568.2
PP tok/s
15.2
TG tok/s
14417
TTFT (ms)
10.5
Peak mem (GB)
Hardware
Chip M5 (8c)
Memory 16 GB
GPU Cores 8
Software
oMLX v0.3.4
macOS macOS 26.4
Context 8,192
Performance by Context Length
Context PP tok/s TG tok/s Peak Mem
1k 520.2 15.9 9.5 GB view
4k 572.2 15.6 10.2 GB view
8k 568.2 15.2 10.5 GB current
16k 524.2 14.6 11.3 GB view
Batching Results
Batch Size TG tok/s Speedup
15.9 1.00×
27.9 1.75×
28.5 1.79×