← Back to community benchmarks
Qwen3.6-27B-oQ4-mtp
Performance
32k
tokens
384.9
PP tok/s
39.0
TG tok/s
85131
TTFT (ms)
23.7
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
32,768
Performance by Context Length
| Context | PP tok/s | TG tok/s | Peak Mem | |
|---|---|---|---|---|
| 1k | 394.0 | 49.9 | 16.8 GB | view |
| 4k | 422.5 | 47.6 | 18.2 GB | view |
| 8k | 421.5 | 46.1 | 19.2 GB | view |
| 16k | 410.5 | 43.7 | 20.7 GB | view |
| 32k | 384.9 | 39.0 | 23.7 GB | current |
Batching Results
| Batch Size | TG tok/s | Speedup |
|---|---|---|
| 1× | 49.9 | 1.00× |
| 2× | 62.6 | 1.25× |
| 4× | 81.0 | 1.62× |