← Back to community benchmarks
Qwen3.6-35B-A3B-NSC-ACE-SABER-6bit-MTPLX-Optimized-Speed
Performance
195k
tokens
165.9
PP tok/s
23.2
TG tok/s
1205610
TTFT (ms)
34.9
Peak mem (GB)
Hardware
Chip
M2 Max (38c)
Memory
64 GB
GPU Cores
38
Software
oMLX
v0.5.1
macOS
macOS 26.5.2
Context
200,000
Performance by Context Length
| Context | PP tok/s | TG tok/s | Peak Mem | |
|---|---|---|---|---|
| 1k | 597.6 | 56.8 | 27.3 GB | view |
| 4k | 520.8 | 60.4 | 28.1 GB | view |
| 8k | 572.1 | 59.7 | 28.4 GB | view |
| 16k | 556.5 | 56.2 | 28.7 GB | view |
| 32k | 488.1 | 52.1 | 29.3 GB | view |
| 64k | 362.5 | 43.9 | 29.9 GB | view |
| 128k | 231.4 | 34.7 | 32.0 GB | view |
| 195k | 165.9 | 23.2 | 34.9 GB | current |
Batching Results
| Batch Size | TG tok/s | Speedup |
|---|---|---|
| 1× | 56.8 | 1.00× |
| 2× | 63.9 | 1.12× |
| 4× | 101.3 | 1.78× |
| 8× | 126.5 | 2.23× |