← Back to community benchmarks
Qwen3.6-27B-uncensored-heretic-v2-Native-MTP-Preserved-oQ8
Performance
16k
tokens
698.1
PP tok/s
16.4
TG tok/s
23471
TTFT (ms)
31.7
Peak mem (GB)
Hardware
Chip
M5 Max (40c)
Memory
128 GB
GPU Cores
40
Software
oMLX
v0.4.4
macOS
macOS 26.5.1
Context
16,384
Performance by Context Length
| Context | PP tok/s | TG tok/s | Peak Mem | |
|---|---|---|---|---|
| 1k | 590.7 | 18.2 | 28.3 GB | view |
| 4k | 796.7 | 18.0 | 29.8 GB | view |
| 8k | 767.4 | 17.8 | 30.4 GB | view |
| 16k | 698.1 | 16.4 | 31.7 GB | current |
| 32k | 586.4 | 15.4 | 34.2 GB | view |
| 64k | 433.3 | 15.2 | 39.2 GB | view |
Batching Results
| Batch Size | TG tok/s | Speedup |
|---|---|---|
| 1× | 18.2 | 1.00× |
| 2× | 30.7 | 1.69× |
| 4× | 40.1 | 2.20× |
| 8× | 46.9 | 2.58× |