← Back to community benchmarks
Qwen3.6-35B-A3Bmxfp8
Performance
1k
tokens
1,541
PP tok/s
72.8
TG tok/s
665
TTFT (ms)
34.4
Peak mem (GB)
Hardware
Chip
M3 Ultra (80c)
Memory
512 GB
GPU Cores
80
Software
oMLX
v0.3.5
macOS
macOS 26.4.1
Context
1,024
Performance by Context Length
| Context | PP tok/s | TG tok/s | Peak Mem | |
|---|---|---|---|---|
| 1k | 1,541 | 72.8 | 34.4 GB | current |
| 4k | 2,324 | 69.5 | 35.2 GB | view |
| 8k | 2,449 | 66.6 | 35.5 GB | view |
| 16k | 2,371 | 64.3 | 36.1 GB | view |
| 32k | 2,071 | 57.1 | 37.5 GB | view |
| 64k | 1,557 | 45.8 | 40.2 GB | view |
Batching Results
| Batch Size | TG tok/s | Speedup |
|---|---|---|
| 1× | 72.8 | 1.00× |
| 2× | 120.0 | 1.65× |
| 4× | 180.8 | 2.48× |