← Back to community benchmarks
Qwen3.5-27B-heretic
Performance
8k
tokens
181.5
PP tok/s
18.8
TG tok/s
45125
TTFT (ms)
29.9
Peak mem (GB)
Hardware
Chip
M2 Ultra (60c)
Memory
64 GB
GPU Cores
60
Software
oMLX
v0.2.9
macOS
macOS 26.3.1
Context
8,192
Performance by Context Length
| Context | PP tok/s | TG tok/s | Peak Mem | |
|---|---|---|---|---|
| 1k | 179.9 | 20.5 | 29.1 GB | view |
| 4k | 182.2 | 19.7 | 29.5 GB | view |
| 8k | 181.5 | 18.8 | 29.9 GB | current |
| 16k | 177.9 | 17.4 | 30.8 GB | view |
| 32k | 170.6 | 14.7 | 32.6 GB | view |
| 64k | 155.5 | 7.6 | 36.2 GB | view |
Batching Results
| Batch Size | TG tok/s | Speedup |
|---|---|---|
| 1× | 20.5 | 1.00× |
| 2× | 33.7 | 1.64× |
| 4× | 67.3 | 3.28× |
| 8× | 134.4 | 6.56× |