← Back to community benchmarks
Qwen3.6-35B-A3B-Heretic
Performance
1k
tokens
1,830
PP tok/s
71.7
TG tok/s
560
TTFT (ms)
65.6
Peak mem (GB)
Hardware
Chip
M5 Max (40c)
Memory
128 GB
GPU Cores
40
Software
oMLX
v0.4.2.dev3
macOS
macOS 26.5.1
Context
1,024
Performance by Context Length
| Context | PP tok/s | TG tok/s | Peak Mem | |
|---|---|---|---|---|
| 1k | 1,830 | 71.7 | 65.6 GB | current |
| 4k | 3,185 | 70.6 | 66.4 GB | view |
| 8k | 3,693 | 70.3 | 66.7 GB | view |
| 16k | 3,432 | 68.2 | 67.4 GB | view |
| 32k | 2,964 | 63.9 | 68.7 GB | view |
| 64k | 2,252 | 57.1 | 71.3 GB | view |
| 128k | 1,491 | 34.9 | 76.8 GB | view |
| 195k | 1,089 | 39.8 | 82.5 GB | view |
Batching Results
| Batch Size | TG tok/s | Speedup |
|---|---|---|
| 1× | 71.7 | 1.00× |
| 2× | 66.1 | 0.92× |
| 4× | 131.8 | 1.84× |
| 8× | 267.4 | 3.73× |