← Back to community benchmarks
Qwen3.6-27B-uncensored-heretic-v2-Native-MTP-Preserved-oQ4
Performance
32k
tokens
663.8
PP tok/s
34.1
TG tok/s
49364
TTFT (ms)
23.7
Peak mem (GB)
Hardware
Chip
M5 Max (40c)
Memory
128 GB
GPU Cores
40
Software
oMLX
v0.4.1
macOS
macOS 26.5.1
Context
32,768
Performance by Context Length
| Context | PP tok/s | TG tok/s | Peak Mem | |
|---|---|---|---|---|
| 1k | 760.7 | 39.8 | 16.8 GB | view |
| 4k | 835.9 | 40.4 | 18.2 GB | view |
| 8k | 807.2 | 39.6 | 19.2 GB | view |
| 16k | 757.1 | 36.1 | 20.7 GB | view |
| 32k | 663.8 | 34.1 | 23.7 GB | current |
Batching Results
| Batch Size | TG tok/s | Speedup |
|---|---|---|
| 1× | 39.8 | 1.00× |
| 2× | 45.1 | 1.13× |
| 4× | 57.4 | 1.44× |