← Back to community benchmarks
Qwen3.5-35B-A3B
Performance
16k
tokens
642.8
PP tok/s
62.7
TG tok/s
25489
TTFT (ms)
20.8
Peak mem (GB)
Hardware
Chip
M4 Pro (16c)
Memory
64 GB
GPU Cores
16
Software
oMLX
v0.2.11
macOS
macOS 26.3.1
Context
16,384
Performance by Context Length
| Context | PP tok/s | TG tok/s | Peak Mem | |
|---|---|---|---|---|
| 1k | 678.6 | 74.1 | 20.1 GB | view |
| 4k | 695.9 | 74.3 | 20.2 GB | view |
| 8k | 667.8 | 69.8 | 20.4 GB | view |
| 16k | 642.8 | 62.7 | 20.8 GB | current |
| 32k | 551.2 | 44.1 | 21.6 GB | view |
| 64k | 445.5 | 27.8 | 23.3 GB | view |
| 128k | 303.3 | 16.9 | 26.7 GB | view |
| 195k | 217.8 | 9.2 | 30.3 GB | view |
Batching Results
| Batch Size | TG tok/s | Speedup |
|---|---|---|
| 1× | 74.1 | 1.00× |
| 2× | 170.9 | 2.31× |
| 4× | 341.0 | 4.60× |
| 8× | 684.2 | 9.23× |