← Back to community benchmarks
gpt-oss-20b-MXFP4-Q8
Performance
16k
tokens
2,257
PP tok/s
65.7
TG tok/s
7261
TTFT (ms)
12.3
Peak mem (GB)
Hardware
Chip
M5 Pro (20c)
Memory
64 GB
GPU Cores
20
Software
oMLX
v0.5.1
macOS
macOS 26.5.2
Context
16,384
Performance by Context Length
| Context | PP tok/s | TG tok/s | Peak Mem | |
|---|---|---|---|---|
| 1k | 1,106 | 82.0 | 11.8 GB | view |
| 4k | 2,281 | 78.2 | 11.8 GB | view |
| 8k | 2,359 | 72.9 | 12.1 GB | view |
| 16k | 2,257 | 65.7 | 12.3 GB | current |
| 32k | 2,018 | 55.3 | 12.8 GB | view |
| 64k | 1,584 | 32.2 | 14.0 GB | view |
| 128k | 1,137 | 28.4 | 16.3 GB | view |
| 195k | 852.5 | 20.4 | 19.5 GB | view |
Batching Results
| Batch Size | TG tok/s | Speedup |
|---|---|---|
| 1× | 82.0 | 1.00× |
| 2× | 124.4 | 1.52× |
| 4× | 156.2 | 1.90× |
| 8× | 298.6 | 3.64× |