← Back to community benchmarks
Qwen3.5-27B
Performance
32k
tokens
203.4
PP tok/s
15.2
TG tok/s
161128
TTFT (ms)
32.6
Peak mem (GB)
Hardware
Chip
M2 Ultra (76c)
Memory
192 GB
GPU Cores
76
Software
oMLX
v0.2.6
macOS
macOS 15.0
Context
32,768
Performance by Context Length
| Context | PP tok/s | TG tok/s | Peak Mem | |
|---|---|---|---|---|
| 1k | 220.8 | 20.4 | 29.1 GB | view |
| 4k | 224.2 | 19.3 | 29.5 GB | view |
| 8k | 222.6 | 18.8 | 29.9 GB | view |
| 16k | 218.2 | 17.6 | 30.8 GB | view |
| 32k | 203.4 | 15.2 | 32.6 GB | current |
| 64k | 144.2 | 12.2 | 36.2 GB | view |
Batching Results
| Batch Size | TG tok/s | Speedup |
|---|---|---|
| 1× | 20.4 | 1.00× |
| 2× | 36.6 | 1.79× |
| 4× | 50.6 | 2.48× |
| 8× | 61.1 | 3.00× |