← Back to community benchmarks

gpt-oss-20b-MXFP4-Q8

M3 (10c) · 24 GB · 4bit · 2026-03-17
Performance
8k
tokens
318.4
PP tok/s
26.4
TG tok/s
25727
TTFT (ms)
11.8
Peak mem (GB)
Hardware
Chip M3 (10c)
Memory 24 GB
GPU Cores 10
Software
oMLX v0.2.17
macOS macOS 15.7.3
Context 8,192
Performance by Context Length
Context PP tok/s TG tok/s Peak Mem
1k 320.9 31.4 11.7 GB view
4k 330.3 29.0 11.8 GB view
8k 318.4 26.4 11.8 GB current
16k 294.1 4.6 12.0 GB view
64k 196.4 11.6 13.2 GB view
128k 135.3 7.0 14.8 GB view
195k 101.6 3.6 16.6 GB view
Batching Results
Batch Size TG tok/s Speedup
31.4 1.00×
41.0 1.31×
47.4 1.51×
46.9 1.49×