← Back to community benchmarks
Qwen3.6-27B-oQ4-mtp
Performance
64k
tokens
298.1
PP tok/s
16.7
TG tok/s
219830
TTFT (ms)
29.7
Peak mem (GB)
Hardware
Chip
M5 Pro (20c)
Memory
64 GB
GPU Cores
20
Software
oMLX
v0.3.9.dev2
macOS
macOS 26.4.1
Context
65,536
Performance by Context Length
| Context | PP tok/s | TG tok/s | Peak Mem | |
|---|---|---|---|---|
| 1k | 423.5 | 25.5 | 16.7 GB | view |
| 4k | 457.9 | 25.1 | 18.2 GB | view |
| 8k | 428.6 | 24.0 | 19.2 GB | view |
| 16k | 417.4 | 22.5 | 20.7 GB | view |
| 32k | 371.6 | 21.0 | 23.7 GB | view |
| 64k | 298.1 | 16.7 | 29.7 GB | current |
Batching Results
| Batch Size | TG tok/s | Speedup |
|---|---|---|
| 1× | 25.5 | 1.00× |
| 2× | 29.3 | 1.15× |
| 4× | 35.3 | 1.38× |
| 8× | 36.8 | 1.44× |