← Back to community benchmarks
Qwen3.6-27B-oQ4-fp16-mtp
Performance
32k
tokens
272.9
PP tok/s
29.4
TG tok/s
120064
TTFT (ms)
23.7
Peak mem (GB)
Hardware
Chip
M2 Ultra (60c)
Memory
64 GB
GPU Cores
60
Software
oMLX
v0.3.9rc1
macOS
macOS 26.3.1
Context
32,768
Performance by Context Length
| Context | PP tok/s | TG tok/s | Peak Mem | |
|---|---|---|---|---|
| 1k | 276.0 | 34.8 | 16.7 GB | view |
| 4k | 296.1 | 36.4 | 18.2 GB | view |
| 8k | 295.4 | 32.8 | 19.2 GB | view |
| 16k | 288.2 | 33.1 | 20.7 GB | view |
| 32k | 272.9 | 29.4 | 23.7 GB | current |
| 64k | 229.9 | 24.6 | 29.7 GB | view |
Batching Results
| Batch Size | TG tok/s | Speedup |
|---|---|---|
| 1× | 34.8 | 1.00× |
| 2× | 46.0 | 1.32× |
| 4× | 53.5 | 1.54× |
| 8× | 59.3 | 1.70× |