← Back to community benchmarks

gpt-oss-20b-MXFP4-Q8

M5 Max (40c) · 128 GB · 4bit · 2026-05-01
Performance
8k
tokens
3,981
PP tok/s
108.3
TG tok/s
2058
TTFT (ms)
12.0
Peak mem (GB)
Hardware
Chip M5 Max (40c)
Memory 128 GB
GPU Cores 40
Software
oMLX v0.3.8
macOS macOS 26.4.1
Context 8,192
Performance by Context Length
Context PP tok/s TG tok/s Peak Mem
1k 1,097 123.4 11.8 GB view
4k 3,385 114.3 11.8 GB view
8k 3,981 108.3 12.0 GB current
16k 3,966 26.7 12.3 GB view
Batching Results
Batch Size TG tok/s Speedup
123.4 1.00×
176.4 1.43×
223.6 1.81×