← Back to community benchmarks

Qwen3.5-9B-OmniCoder-Claude-Polaris-qx86-hi

M2 Max (38c) · 96 GB · 8bit · 2026-03-16

Performance

4k

tokens

402.5

PP tok/s

39.7

TG tok/s

10177

TTFT (ms)

10.2

Peak mem (GB)

Hardware

Chip M2 Max (38c)

Memory 96 GB

GPU Cores 38

Software

oMLX v0.2.13

macOS macOS 26.3.1

Context 4,096

Performance by Context Length

Context	PP tok/s	TG tok/s	Peak Mem
1k	395.5	40.4	10.0 GB	view
4k	402.5	39.7	10.2 GB	current

Batching Results

Batch Size	TG tok/s	Speedup
1×	40.4	1.00×
2×	66.7	1.65×
4×	79.1	1.96×