← Back to community benchmarks
Qwen3.5-122B-A10B9bit
Performance
1k
tokens
664.7
PP tok/s
42.1
TG tok/s
1541
TTFT (ms)
129.5
Peak mem (GB)
Hardware
Chip
M3 Ultra (80c)
Memory
512 GB
GPU Cores
80
Software
oMLX
v0.2.13
macOS
macOS 26.3.1
Context
1,024
Performance by Context Length
| Context | PP tok/s | TG tok/s | Peak Mem | |
|---|---|---|---|---|
| 1k | 664.7 | 42.1 | 129.5 GB | current |
| 4k | 855.2 | 40.8 | 129.8 GB | view |
| 8k | 872.7 | 39.3 | 130.1 GB | view |
| 16k | 829.2 | 36.6 | 130.8 GB | view |
| 32k | 738.9 | 32.1 | 132.2 GB | view |
| 64k | 581.5 | 26.4 | 134.9 GB | view |
Batching Results
| Batch Size | TG tok/s | Speedup |
|---|---|---|
| 1× | 42.1 | 1.00× |
| 2× | 64.7 | 1.54× |
| 4× | 88.1 | 2.09× |