← Back to community benchmarks
Qwen3.6-27B-FP8-source-oQ4-mtp
Performance
64k
tokens
139.6
PP tok/s
18.9
TG tok/s
469428
TTFT (ms)
29.9
Peak mem (GB)
Hardware
Chip
M4 Max (40c)
Memory
128 GB
GPU Cores
40
Software
oMLX
v0.3.9.dev1
macOS
macOS 26.4.1
Context
65,536
Performance by Context Length
| Context | PP tok/s | TG tok/s | Peak Mem | |
|---|---|---|---|---|
| 1k | 236.5 | 38.6 | 16.9 GB | view |
| 4k | 184.8 | 27.6 | 18.4 GB | view |
| 8k | 179.4 | 31.9 | 19.4 GB | view |
| 16k | 159.6 | 22.4 | 20.9 GB | view |
| 32k | 154.2 | 22.8 | 23.9 GB | view |
| 64k | 139.6 | 18.9 | 29.9 GB | current |
Batching Results
| Batch Size | TG tok/s | Speedup |
|---|---|---|
| 1× | 38.6 | 1.00× |
| 2× | 34.9 | 0.90× |
| 4× | 40.3 | 1.04× |