← Back to community benchmarks

Qwen3.5-27B-heretic

M1 Ultra (64c) · 128 GB · 8bit · 2026-03-10

Performance

8k

tokens

159.4

PP tok/s

16.0

TG tok/s

51378

TTFT (ms)

29.9

Peak mem (GB)

Hardware

Chip M1 Ultra (64c)

Memory 128 GB

GPU Cores 64

Software

oMLX v0.2.6

macOS macOS 26.4

Context 8,192

Performance by Context Length

Context	PP tok/s	TG tok/s	Peak Mem
1k	159.1	17.4	29.1 GB	view
4k	160.5	16.7	29.5 GB	view
8k	159.4	16.0	29.9 GB	current

Batching Results

Batch Size	TG tok/s	Speedup
1×	17.4	1.00×
2×	30.5	1.75×
4×	38.2	2.20×