← Back to community benchmarks
Qwen3.6-35B-A3Bmixed-9bit
Performance
16k
tokens
2,048
PP tok/s
79.3
TG tok/s
7999
TTFT (ms)
40.0
Peak mem (GB)
Hardware
Chip
M3 Ultra (60c)
Memory
96 GB
GPU Cores
60
Software
oMLX
v0.5.3
macOS
macOS 26.5.2
Context
16,384
Performance by Context Length
| Context | PP tok/s | TG tok/s | Peak Mem | |
|---|---|---|---|---|
| 1k | 2,135 | 86.1 | 37.9 GB | view |
| 4k | 2,326 | 84.9 | 38.7 GB | view |
| 8k | 2,230 | 83.4 | 39.2 GB | view |
| 16k | 2,048 | 79.3 | 40.0 GB | current |
| 32k | 1,752 | 73.6 | 41.4 GB | view |
| 64k | 1,334 | 64.4 | 44.4 GB | view |
| 128k | 856.0 | 52.1 | 50.3 GB | view |
| 195k | 615.8 | 43.7 | 57.3 GB | view |
Batching Results
| Batch Size | TG tok/s | Speedup |
|---|---|---|
| 1× | 86.1 | 1.00× |
| 2× | 141.4 | 1.64× |
| 4× | 226.4 | 2.63× |
| 8× | 346.6 | 4.03× |