← Back to community benchmarks
Qwen3.5-35B-A3B
Performance
32k
tokens
1,703
PP tok/s
62.3
TG tok/s
19247
TTFT (ms)
21.6
Peak mem (GB)
Hardware
Chip
M5 Pro (20c)
Memory
64 GB
GPU Cores
20
Software
oMLX
v0.2.12
macOS
macOS 26.3.1
Context
32,768
Performance by Context Length
| Context | PP tok/s | TG tok/s | Peak Mem | |
|---|---|---|---|---|
| 1k | 1,599 | 98.4 | 20.1 GB | view |
| 4k | 2,095 | 94.9 | 20.2 GB | view |
| 8k | 2,118 | 90.3 | 20.4 GB | view |
| 16k | 1,981 | 76.0 | 20.8 GB | view |
| 32k | 1,703 | 62.3 | 21.6 GB | current |
| 64k | 1,278 | 41.6 | 23.3 GB | view |
| 128k | 832.6 | 15.5 | 26.7 GB | view |
| 195k | 604.5 | 10.6 | 30.3 GB | view |
Batching Results
| Batch Size | TG tok/s | Speedup |
|---|---|---|
| 1× | 98.4 | 1.00× |
| 2× | 155.3 | 1.58× |
| 4× | 211.2 | 2.15× |
| 8× | 269.1 | 2.73× |