← Back to community benchmarks
Qwen3.6-35B-A3B
Performance
16k
tokens
1,127
PP tok/s
63.4
TG tok/s
14544
TTFT (ms)
21.0
Peak mem (GB)
Hardware
Chip
M2 Ultra (60c)
Memory
64 GB
GPU Cores
60
Software
oMLX
v0.3.6
macOS
macOS 26.4.1
Context
16,384
Performance by Context Length
| Context | PP tok/s | TG tok/s | Peak Mem | |
|---|---|---|---|---|
| 1k | 881.4 | 89.9 | 19.3 GB | view |
| 4k | 1,147 | 84.6 | 20.0 GB | view |
| 8k | 1,165 | 79.1 | 20.4 GB | view |
| 16k | 1,127 | 63.4 | 21.0 GB | current |
| 32k | 1,024 | 54.8 | 22.4 GB | view |
| 64k | 431.9 | 22.9 | 25.0 GB | view |
Batching Results
| Batch Size | TG tok/s | Speedup |
|---|---|---|
| 1× | 89.9 | 1.00× |
| 2× | 86.4 | 0.96× |
| 4× | 120.1 | 1.34× |
| 8× | 126.4 | 1.41× |