← Back to community benchmarks
Qwen3.6-35B-A3B
Performance
8k
tokens
2,698
PP tok/s
95.0
TG tok/s
3037
TTFT (ms)
20.4
Peak mem (GB)
Hardware
Chip
M3 Ultra (80c)
Memory
512 GB
GPU Cores
80
Software
oMLX
v0.3.7
macOS
macOS 26.4.1
Context
8,192
Performance by Context Length
| Context | PP tok/s | TG tok/s | Peak Mem | |
|---|---|---|---|---|
| 1k | 1,769 | 104.5 | 19.3 GB | view |
| 4k | 2,613 | 100.6 | 20.0 GB | view |
| 8k | 2,698 | 95.0 | 20.4 GB | current |
| 16k | 2,567 | 86.9 | 21.0 GB | view |
| 32k | 2,222 | 74.4 | 22.4 GB | view |
| 64k | 1,674 | 58.8 | 25.0 GB | view |
| 128k | 1,047 | 40.9 | 30.4 GB | view |
| 195k | 742.6 | 30.2 | 36.1 GB | view |
Batching Results
| Batch Size | TG tok/s | Speedup |
|---|---|---|
| 1× | 104.5 | 1.00× |
| 2× | 176.9 | 1.69× |
| 4× | 285.1 | 2.73× |