← Back to community benchmarks
Qwen--Qwen3.5-35B-A3B
Performance
4k
tokens
871.0
PP tok/s
2.9
TG tok/s
4703
TTFT (ms)
72.7
Peak mem (GB)
Hardware
Chip
M3 Ultra (80c)
Memory
512 GB
GPU Cores
80
Software
oMLX
v0.2.7
macOS
macOS 26.3.1
Context
4,096
Performance by Context Length
| Context | PP tok/s | TG tok/s | Peak Mem | |
|---|---|---|---|---|
| 1k | 507.6 | 2.9 | 72.6 GB | view |
| 4k | 871.0 | 2.9 | 72.7 GB | current |
| 8k | 970.8 | 3.0 | 72.9 GB | view |
| 16k | 977.6 | 3.0 | 73.2 GB | view |
| 32k | 914.1 | 2.9 | 73.8 GB | view |
| 64k | 707.2 | 2.9 | 76.6 GB | view |
Batching Results
| Batch Size | TG tok/s | Speedup |
|---|---|---|
| 1× | 2.9 | 1.00× |
| 2× | 5.8 | 2.00× |
| 4× | 11.3 | 3.90× |
| 8× | 18.3 | 6.31× |