← Back to community benchmarks
Qwen3.6-35B-A3B-mxfp8
Performance
128k
tokens
1,036
PP tok/s
36.2
TG tok/s
126471
TTFT (ms)
45.5
Peak mem (GB)
Hardware
Chip
M3 Ultra (80c)
Memory
512 GB
GPU Cores
80
Software
oMLX
v0.3.6
macOS
macOS 26.4.1
Context
131,072
Performance by Context Length
| Context | PP tok/s | TG tok/s | Peak Mem | |
|---|---|---|---|---|
| 1k | 1,775 | 83.0 | 34.4 GB | view |
| 4k | 2,629 | 80.9 | 35.2 GB | view |
| 8k | 2,739 | 77.7 | 35.5 GB | view |
| 16k | 2,604 | 72.0 | 36.1 GB | view |
| 32k | 2,244 | 63.3 | 37.5 GB | view |
| 64k | 1,677 | 50.5 | 40.2 GB | view |
| 128k | 1,036 | 36.2 | 45.5 GB | current |
| 195k | 732.9 | 28.0 | 51.2 GB | view |
Batching Results
| Batch Size | TG tok/s | Speedup |
|---|---|---|
| 1× | 83.0 | 1.00× |
| 2× | 145.9 | 1.76× |
| 4× | 205.7 | 2.48× |
| 8× | 281.5 | 3.39× |