← Back to community benchmarks
Qwen3.5-27B
Performance
64k
tokens
149.2
PP tok/s
3.8
TG tok/s
439316
TTFT (ms)
68.7
Peak mem (GB)
Hardware
Chip
M3 Max (40c)
Memory
128 GB
GPU Cores
40
Software
oMLX
v0.2.11
macOS
macOS 15.7.4
Context
65,536
Performance by Context Length
| Context | PP tok/s | TG tok/s | Peak Mem | |
|---|---|---|---|---|
| 1k | 211.6 | 12.9 | 29.1 GB | view |
| 4k | 210.6 | 12.6 | 29.5 GB | view |
| 8k | 187.4 | 11.8 | 29.9 GB | view |
| 16k | 189.9 | 10.2 | 30.8 GB | view |
| 32k | 173.0 | 7.9 | 32.6 GB | view |
| 64k | 149.2 | 3.8 | 68.7 GB | current |
| 128k | 118.0 | 4.5 | 78.6 GB | view |
| 195k | 97.2 | 2.4 | 86.3 GB | view |
Batching Results
| Batch Size | TG tok/s | Speedup |
|---|---|---|
| 1× | 12.9 | 1.00× |
| 2× | 22.9 | 1.78× |
| 4× | 29.3 | 2.27× |
| 8× | 29.8 | 2.31× |