← Back to community benchmarks
Qwen3___5-35B-A3B
Performance
32k
tokens
2,098
PP tok/s
58.3
TG tok/s
15622
TTFT (ms)
37.8
Peak mem (GB)
Hardware
Chip
M3 Ultra (80c)
Memory
512 GB
GPU Cores
80
Software
oMLX
v0.2.6
macOS
macOS 26.3.1
Context
32,768
Performance by Context Length
| Context | PP tok/s | TG tok/s | Peak Mem | |
|---|---|---|---|---|
| 1k | 1,678 | 76.7 | 36.2 GB | view |
| 4k | 2,415 | 74.3 | 36.3 GB | view |
| 8k | 2,621 | 71.8 | 36.5 GB | view |
| 16k | 2,429 | 67.1 | 37.0 GB | view |
| 32k | 2,098 | 58.3 | 37.8 GB | current |
| 64k | 1,608 | 48.3 | 39.5 GB | view |
Batching Results
| Batch Size | TG tok/s | Speedup |
|---|---|---|
| 1× | 76.7 | 1.00× |
| 2× | 134.3 | 1.75× |
| 4× | 211.8 | 2.76× |