← Back to community benchmarks
Qwen3.5-27BMTP
Performance
32k
tokens
83.7
PP tok/s
14.1
TG tok/s
391456
TTFT (ms)
23.1
Peak mem (GB)
Hardware
Chip
M1 Max (32c)
Memory
64 GB
GPU Cores
32
Software
oMLX
v0.3.9.dev1
macOS
macOS 15.7.4
Context
32,768
Performance by Context Length
| Context | PP tok/s | TG tok/s | Peak Mem | |
|---|---|---|---|---|
| 1k | 86.9 | 18.1 | 16.1 GB | view |
| 4k | 88.8 | 17.7 | 17.6 GB | view |
| 8k | 88.5 | 17.4 | 18.6 GB | view |
| 16k | 86.9 | 16.1 | 20.1 GB | view |
| 32k | 83.7 | 14.1 | 23.1 GB | current |
| 64k | 76.0 | 11.7 | 29.1 GB | view |
| 128k | 61.9 | 8.7 | 41.5 GB | view |
Batching Results
| Batch Size | TG tok/s | Speedup |
|---|---|---|
| 1× | 18.1 | 1.00× |
| 2× | 19.2 | 1.06× |
| 4× | 23.7 | 1.31× |