← Back to community benchmarks
Qwen3.6-35B-A3B-UD
Performance
128k
tokens
985.4
PP tok/s
48.2
TG tok/s
133019
TTFT (ms)
32.9
Peak mem (GB)
Hardware
Chip
M3 Ultra (80c)
Memory
512 GB
GPU Cores
80
Software
oMLX
v0.3.8
macOS
macOS 15.6
Context
131,072
Performance by Context Length
| Context | PP tok/s | TG tok/s | Peak Mem | |
|---|---|---|---|---|
| 1k | 1,738 | 88.0 | 20.4 GB | view |
| 4k | 2,563 | 86.4 | 21.2 GB | view |
| 8k | 2,652 | 82.4 | 21.7 GB | view |
| 16k | 2,514 | 77.5 | 22.4 GB | view |
| 32k | 2,138 | 70.9 | 23.9 GB | view |
| 64k | 1,587 | 62.7 | 26.9 GB | view |
| 128k | 985.4 | 48.2 | 32.9 GB | current |
Batching Results
| Batch Size | TG tok/s | Speedup |
|---|---|---|
| 1× | 88.0 | 1.00× |
| 2× | 135.1 | 1.54× |
| 4× | 223.6 | 2.54× |