← Back to community benchmarks
Qwen3.6-27B-oQ8-mtp
Performance
16k
tokens
409.4
PP tok/s
28.5
TG tok/s
40020
TTFT (ms)
32.8
Peak mem (GB)
Hardware
Chip
M3 Ultra (80c)
Memory
512 GB
GPU Cores
80
Software
oMLX
v0.3.9.dev2
macOS
macOS 26.5
Context
16,384
Performance by Context Length
| Context | PP tok/s | TG tok/s | Peak Mem | |
|---|---|---|---|---|
| 1k | 387.7 | 32.4 | 28.8 GB | view |
| 4k | 420.8 | 31.4 | 30.3 GB | view |
| 8k | 420.1 | 30.5 | 31.3 GB | view |
| 16k | 409.4 | 28.5 | 32.8 GB | current |
| 32k | 384.7 | 27.3 | 35.8 GB | view |
| 64k | 331.2 | 24.0 | 41.8 GB | view |
| 128k | 256.2 | 19.6 | 54.2 GB | view |
Batching Results
| Batch Size | TG tok/s | Speedup |
|---|---|---|
| 1× | 32.4 | 1.00× |
| 2× | 38.7 | 1.19× |
| 4× | 58.1 | 1.79× |