← Back to community benchmarks
Qwen3.6-27B-oQ8-mtp
Performance
16k
tokens
409.5
PP tok/s
31.3
TG tok/s
40005
TTFT (ms)
32.8
Peak mem (GB)
Hardware
Chip
M3 Ultra (80c)
Memory
96 GB
GPU Cores
80
Software
oMLX
v0.3.9.dev1
macOS
macOS 26.5
Context
16,384
Performance by Context Length
| Context | PP tok/s | TG tok/s | Peak Mem | |
|---|---|---|---|---|
| 1k | 388.4 | 30.8 | 28.8 GB | view |
| 4k | 420.9 | 33.6 | 30.3 GB | view |
| 8k | 420.3 | 32.7 | 31.3 GB | view |
| 16k | 409.5 | 31.3 | 32.8 GB | current |
| 32k | 385.1 | 29.1 | 35.8 GB | view |
| 64k | 335.0 | 25.6 | 41.8 GB | view |
Batching Results
| Batch Size | TG tok/s | Speedup |
|---|---|---|
| 1× | 30.8 | 1.00× |
| 2× | 40.2 | 1.31× |
| 4× | 61.8 | 2.01× |