← Back to community benchmarks
Qwen3.6-35B-A3B-oQ3-mtp
Performance
16k
tokens
2,605
PP tok/s
76.8
TG tok/s
6290
TTFT (ms)
18.3
Peak mem (GB)
Hardware
Chip
M3 Ultra (80c)
Memory
512 GB
GPU Cores
80
Software
oMLX
v0.4.2rc1
macOS
macOS 26.3
Context
16,384
Performance by Context Length
| Context | PP tok/s | TG tok/s | Peak Mem | |
|---|---|---|---|---|
| 1k | 2,171 | 72.1 | 16.3 GB | view |
| 4k | 2,783 | 82.7 | 17.1 GB | view |
| 8k | 2,791 | 80.2 | 17.6 GB | view |
| 16k | 2,605 | 76.8 | 18.3 GB | current |
| 32k | 2,241 | 70.9 | 19.8 GB | view |
| 64k | 1,676 | 61.4 | 22.8 GB | view |
Batching Results
| Batch Size | TG tok/s | Speedup |
|---|---|---|
| 1× | 72.1 | 1.00× |
| 2× | 161.1 | 2.23× |
| 4× | 230.8 | 3.20× |