← Back to community benchmarks
Qwen3.5-9B
Performance
64k
tokens
1,016
PP tok/s
27.8
TG tok/s
64508
TTFT (ms)
10.8
Peak mem (GB)
Hardware
Chip
M3 Ultra (80c)
Memory
512 GB
GPU Cores
80
Software
oMLX
v0.2.8
macOS
macOS 15.6
Context
65,536
Performance by Context Length
| Context | PP tok/s | TG tok/s | Peak Mem | |
|---|---|---|---|---|
| 1k | 1,376 | 78.8 | 6.7 GB | view |
| 4k | 1,431 | 83.3 | 7.0 GB | view |
| 8k | 1,420 | 88.5 | 7.1 GB | view |
| 16k | 1,344 | 47.0 | 7.6 GB | view |
| 32k | 1,231 | 40.5 | 8.7 GB | view |
| 64k | 1,016 | 27.8 | 10.8 GB | current |
Batching Results
| Batch Size | TG tok/s | Speedup |
|---|---|---|
| 1× | 78.8 | 1.00× |
| 2× | 109.2 | 1.39× |
| 4× | 209.0 | 2.65× |
| 8× | 269.6 | 3.42× |