← Back to community benchmarks
Qwen3.6-27B
Performance
64k
tokens
1,117
PP tok/s
27.7
TG tok/s
58677
TTFT (ms)
23.3
Peak mem (GB)
Hardware
Chip
M3 Ultra (60c)
Memory
96 GB
GPU Cores
60
Software
oMLX
v0.3.7rc2
macOS
macOS 26.4.1
Context
65,536
Performance by Context Length
| Context | PP tok/s | TG tok/s | Peak Mem | |
|---|---|---|---|---|
| 1k | 304.6 | 39.1 | 16.4 GB | view |
| 4k | 324.0 | 37.2 | 17.8 GB | view |
| 8k | 322.0 | 35.5 | 18.5 GB | view |
| 16k | 1,229 | 30.4 | 20.2 GB | view |
| 32k | 1,205 | 29.5 | 21.4 GB | view |
| 64k | 1,117 | 27.7 | 23.3 GB | current |
| 128k | 952.4 | 25.4 | 27.0 GB | view |
| 195k | 803.6 | 23.2 | 31.4 GB | view |
Batching Results
| Batch Size | TG tok/s | Speedup |
|---|---|---|
| 1× | 39.1 | 1.00× |
| 2× | 63.4 | 1.62× |