← Back to community benchmarks
Qwen3.6-27B-oQ4-fp16-mtp
Performance
64k
tokens
373.2
PP tok/s
20.3
TG tok/s
175589
TTFT (ms)
29.7
Peak mem (GB)
Hardware
Chip
M5 Max (40c)
Memory
128 GB
GPU Cores
40
Software
oMLX
v0.3.9
macOS
macOS 26.4
Context
65,536
Performance by Context Length
| Context | PP tok/s | TG tok/s | Peak Mem | |
|---|---|---|---|---|
| 1k | 765.4 | 44.1 | 16.7 GB | view |
| 4k | 790.6 | 42.9 | 18.2 GB | view |
| 8k | 725.5 | 41.2 | 19.2 GB | view |
| 16k | 656.0 | 38.8 | 20.7 GB | view |
| 32k | 620.5 | 28.0 | 23.7 GB | view |
| 64k | 373.2 | 20.3 | 29.7 GB | current |
Batching Results
| Batch Size | TG tok/s | Speedup |
|---|---|---|
| 1× | 44.1 | 1.00× |
| 2× | 34.9 | 0.79× |
| 4× | 42.0 | 0.95× |
| 8× | 46.5 | 1.05× |