← Back to community benchmarks
Qwen3.6-27B-oQ8-mtp
Performance
32k
tokens
701.7
PP tok/s
24.8
TG tok/s
46701
TTFT (ms)
37.9
Peak mem (GB)
Hardware
Chip
M5 Max (40c)
Memory
128 GB
GPU Cores
40
Software
oMLX
v0.4.1
macOS
macOS 26.6
Context
32,768
Performance by Context Length
| Context | PP tok/s | TG tok/s | Peak Mem | |
|---|---|---|---|---|
| 1k | 703.1 | 19.5 | 30.9 GB | view |
| 4k | 837.8 | 24.6 | 32.4 GB | view |
| 8k | 802.5 | 25.9 | 33.4 GB | view |
| 16k | 766.2 | 26.4 | 34.9 GB | view |
| 32k | 701.7 | 24.8 | 37.9 GB | current |
| 64k | 569.1 | 21.6 | 43.9 GB | view |
| 128k | 400.7 | 17.2 | 56.3 GB | view |
Batching Results
| Batch Size | TG tok/s | Speedup |
|---|---|---|
| 1× | 19.5 | 1.00× |
| 2× | 31.4 | 1.61× |
| 4× | 44.2 | 2.27× |
| 8× | 86.1 | 4.42× |