← Back to community benchmarks
Qwen3.6-27B-oQ4-mtp
Performance
16k
tokens
159.6
PP tok/s
18.8
TG tok/s
102668
TTFT (ms)
20.7
Peak mem (GB)
Hardware
Chip
M4 Max (40c)
Memory
128 GB
GPU Cores
40
Software
oMLX
v0.3.9.dev1
macOS
macOS 26.3.1
Context
16,384
Performance by Context Length
| Context | PP tok/s | TG tok/s | Peak Mem | |
|---|---|---|---|---|
| 1k | 240.1 | 35.7 | 16.8 GB | view |
| 4k | 190.9 | 30.2 | 18.2 GB | view |
| 8k | 188.1 | 33.1 | 19.2 GB | view |
| 16k | 159.6 | 18.8 | 20.7 GB | current |
| 32k | 160.5 | 23.2 | 23.7 GB | view |
| 64k | 145.8 | 18.7 | 29.8 GB | view |
| 128k | 117.6 | 14.5 | 42.1 GB | view |
| 195k | 97.1 | 11.4 | 56.8 GB | view |
Batching Results
| Batch Size | TG tok/s | Speedup |
|---|---|---|
| 1× | 35.7 | 1.00× |
| 2× | 26.1 | 0.73× |
| 4× | 38.2 | 1.07× |
| 8× | 41.1 | 1.15× |