← Back to community benchmarks

gpt-oss-20b-MXFP4-Q8

M3 Ultra (60c) · 96 GB · 4bit · 2026-04-25

Performance

4k

tokens

1,748

PP tok/s

103.5

TG tok/s

2343

TTFT (ms)

11.8

Peak mem (GB)

Hardware

Chip M3 Ultra (60c)

Memory 96 GB

GPU Cores 60

Software

oMLX v0.3.6

macOS macOS 26.3.2

Context 4,096

Performance by Context Length

Context	PP tok/s	TG tok/s	Peak Mem
1k	1,129	112.4	11.8 GB	view
4k	1,748	103.5	11.8 GB	current

Batching Results

Batch Size	TG tok/s	Speedup
1×	112.4	1.00×
2×	155.4	1.38×
4×	199.6	1.78×