← Back to community benchmarks
Qwen36-27B-oQ4-mtp
Performance
16k
tokens
240.0
PP tok/s
34.7
TG tok/s
68264
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 15.7.4
Context
16,384
Performance by Context Length
| Context | PP tok/s | TG tok/s | Peak Mem | |
|---|---|---|---|---|
| 1k | 240.4 | 37.4 | 16.8 GB | view |
| 4k | 248.4 | 38.8 | 18.2 GB | view |
| 8k | 246.7 | 38.1 | 19.2 GB | view |
| 16k | 240.0 | 34.7 | 20.7 GB | current |
| 32k | 226.3 | 32.1 | 23.7 GB | view |
| 64k | 201.7 | 26.0 | 29.8 GB | view |
Batching Results
| Batch Size | TG tok/s | Speedup |
|---|---|---|
| 1× | 37.4 | 1.00× |
| 2× | 49.5 | 1.32× |
| 4× | 56.0 | 1.50× |