← Back to community benchmarks
Qwen3.6-35B-A3B-PARO
Performance
195k
tokens
593.1
PP tok/s
39.1
TG tok/s
337185
TTFT (ms)
23.9
Peak mem (GB)
Hardware
Chip
M3 Ultra (60c)
Memory
256 GB
GPU Cores
60
Software
oMLX
v0.5.2.dev1
macOS
macOS 27.0
Context
200,000
Performance by Context Length
| Context | PP tok/s | TG tok/s | Peak Mem | |
|---|---|---|---|---|
| 1k | 1,823 | 69.3 | 18.6 GB | view |
| 4k | 2,361 | 67.6 | 19.2 GB | view |
| 8k | 2,309 | 67.6 | 19.3 GB | view |
| 16k | 2,048 | 63.4 | 19.4 GB | view |
| 32k | 1,682 | 59.8 | 19.8 GB | view |
| 64k | 1,236 | 54.9 | 20.5 GB | view |
| 128k | 808.6 | 44.9 | 22.1 GB | view |
| 195k | 593.1 | 39.1 | 23.9 GB | current |
Batching Results
| Batch Size | TG tok/s | Speedup |
|---|---|---|
| 1× | 69.3 | 1.00× |
| 2× | 108.2 | 1.56× |