← Back to community benchmarks

DeepSeek-R1-0528-Qwen3-8B

M4 (10c) · 16 GB · 4bit · 2026-04-07

Performance

4k

tokens

210.0

PP tok/s

19.3

TG tok/s

19507

TTFT (ms)

5.4

Peak mem (GB)

Hardware

Chip M4 (10c)

Memory 16 GB

GPU Cores 10

Software

oMLX v0.3.0rc1

macOS macOS 26.3.1

Context 4,096

Performance by Context Length

Context	PP tok/s	TG tok/s	Peak Mem
1k	220.6	21.8	4.9 GB	view
4k	210.0	19.3	5.4 GB	current

Batching Results

Batch Size	TG tok/s	Speedup
1×	21.8	1.00×
2×	42.2	1.94×
4×	52.9	2.43×