← Back to community benchmarks
Qwen3.5-122B-A10B-Text-mxfp4
Performance
32k
tokens
387.4
PP tok/s
27.7
TG tok/s
84578
TTFT (ms)
67.5
Peak mem (GB)
Hardware
Chip
M3 Max (40c)
Memory
128 GB
GPU Cores
40
Software
oMLX
v0.3.4
macOS
macOS 26.4
Context
32,768
Performance by Context Length
| Context | PP tok/s | TG tok/s | Peak Mem | |
|---|---|---|---|---|
| 1k | 431.8 | 44.5 | 62.0 GB | view |
| 4k | 496.2 | 41.7 | 63.4 GB | view |
| 8k | 483.0 | 40.3 | 64.0 GB | view |
| 16k | 457.9 | 36.6 | 65.1 GB | view |
| 32k | 387.4 | 27.7 | 67.5 GB | current |
Batching Results
| Batch Size | TG tok/s | Speedup |
|---|---|---|
| 1× | 44.5 | 1.00× |
| 2× | 67.5 | 1.52× |
| 4× | 91.3 | 2.05× |