← Back to community benchmarks
Qwen3.5-122B-A10B
Performance
8k
tokens
915.4
PP tok/s
47.2
TG tok/s
8949
TTFT (ms)
67.0
Peak mem (GB)
Hardware
Chip
M3 Ultra (80c)
Memory
512 GB
GPU Cores
80
Software
oMLX
v0.2.10
macOS
macOS 15.7.4
Context
8,192
Performance by Context Length
| Context | PP tok/s | TG tok/s | Peak Mem | |
|---|---|---|---|---|
| 1k | 894.1 | 52.4 | 66.3 GB | view |
| 4k | 927.0 | 49.9 | 66.7 GB | view |
| 8k | 915.4 | 47.2 | 67.0 GB | current |
| 16k | 856.4 | 43.1 | 67.7 GB | view |
| 32k | 740.0 | 36.8 | 69.1 GB | view |
Batching Results
| Batch Size | TG tok/s | Speedup |
|---|---|---|
| 1× | 52.4 | 1.00× |
| 2× | 86.6 | 1.65× |
| 4× | 133.9 | 2.56× |
| 8× | 187.1 | 3.57× |