← Back to community benchmarks
Qwen3.5-122B-A10B
Performance
1k
tokens
729.4
PP tok/s
29.1
TG tok/s
1404
TTFT (ms)
229.8
Peak mem (GB)
Hardware
Chip
M3 Ultra (80c)
Memory
512 GB
GPU Cores
80
Software
oMLX
v0.2.11
macOS
macOS 26.3.1
Context
1,024
Performance by Context Length
| Context | PP tok/s | TG tok/s | Peak Mem | |
|---|---|---|---|---|
| 1k | 729.4 | 29.1 | 229.8 GB | current |
| 4k | 840.0 | 28.3 | 230.1 GB | view |
| 8k | 848.2 | 27.6 | 230.5 GB | view |
| 16k | 811.7 | 25.6 | 231.2 GB | view |
| 32k | 719.4 | 22.3 | 232.6 GB | view |
| 64k | 561.9 | 16.1 | 235.3 GB | view |
Batching Results
| Batch Size | TG tok/s | Speedup |
|---|---|---|
| 1× | 29.1 | 1.00× |
| 2× | 32.2 | 1.11× |
| 4× | 48.7 | 1.67× |