← Back to community benchmarks
deepsweet-Qwen3.6-35B-A3BVL-oQ6
Performance
1k
tokens
1,317
PP tok/s
54.5
TG tok/s
778
TTFT (ms)
27.6
Peak mem (GB)
Hardware
Chip
M2 Ultra (60c)
Memory
64 GB
GPU Cores
60
Software
oMLX
v0.4.2rc1
macOS
macOS 15.7.7
Context
1,024
Performance by Context Length
| Context | PP tok/s | TG tok/s | Peak Mem | |
|---|---|---|---|---|
| 1k | 1,317 | 54.5 | 27.6 GB | current |
| 4k | 1,612 | 53.9 | 28.4 GB | view |
| 8k | 1,596 | 52.7 | 28.7 GB | view |
| 16k | 1,482 | 51.2 | 29.4 GB | view |
| 32k | 1,277 | 49.4 | 30.7 GB | view |
| 64k | 839.4 | 31.1 | 33.4 GB | view |
Batching Results
| Batch Size | TG tok/s | Speedup |
|---|---|---|
| 1× | 54.5 | 1.00× |
| 2× | 49.6 | 0.91× |
| 4× | 103.4 | 1.90× |