← Back to community benchmarks
Qwen3.6-27B-oQ4-mtp
Performance
128k
tokens
388.5
PP tok/s
21.9
TG tok/s
337391
TTFT (ms)
42.1
Peak mem (GB)
Hardware
Chip
M5 Max (40c)
Memory
128 GB
GPU Cores
40
Software
oMLX
v0.3.9
macOS
macOS 26.5
Context
131,072
Performance by Context Length
| Context | PP tok/s | TG tok/s | Peak Mem | |
|---|---|---|---|---|
| 1k | 654.0 | 42.9 | 16.8 GB | view |
| 4k | 690.4 | 40.5 | 18.2 GB | view |
| 8k | 717.5 | 39.4 | 19.2 GB | view |
| 16k | 699.5 | 38.6 | 20.7 GB | view |
| 32k | 655.4 | 35.1 | 23.7 GB | view |
| 64k | 546.0 | 28.2 | 29.8 GB | view |
| 128k | 388.5 | 21.9 | 42.1 GB | current |
| 195k | 302.9 | 17.3 | 56.8 GB | view |
Batching Results
| Batch Size | TG tok/s | Speedup |
|---|---|---|
| 1× | 42.9 | 1.00× |
| 2× | 48.9 | 1.14× |
| 4× | 60.6 | 1.41× |
| 8× | 67.4 | 1.57× |