← Back to community benchmarks
Qwen3.5-122B-A10B9bit
Performance
64k
tokens
570.0
PP tok/s
25.5
TG tok/s
114972
TTFT (ms)
140.1
Peak mem (GB)
Hardware
Chip
M3 Ultra (80c)
Memory
512 GB
GPU Cores
80
Software
oMLX
v0.3.0rc1
macOS
macOS 26.1
Context
65,536
Performance by Context Length
| Context | PP tok/s | TG tok/s | Peak Mem | |
|---|---|---|---|---|
| 1k | 691.5 | 38.7 | 129.6 GB | view |
| 4k | 881.6 | 37.6 | 131.0 GB | view |
| 8k | 906.0 | 36.5 | 131.6 GB | view |
| 16k | 858.6 | 34.5 | 132.8 GB | view |
| 32k | 752.2 | 30.5 | 135.2 GB | view |
| 64k | 570.0 | 25.5 | 140.1 GB | current |
Batching Results
| Batch Size | TG tok/s | Speedup |
|---|---|---|
| 1× | 38.7 | 1.00× |
| 2× | 62.2 | 1.61× |
| 4× | 90.1 | 2.33× |
| 8× | 125.6 | 3.25× |